Is the famous ‘paradox of choice’ a myth?
Paul Solman: In 2003, our NewsHour economics crew traipsed to the western outskirts of Philadelphia to rendezvous with Swarthmore psychology professor Barry Schwartz and hear him make the case, at the something-for-everyone King of Prussia Mall, for the thesis around which he’d just written a book, “The Paradox of Choice.”
Watch the segment below.
A decade, a TED talk and a Freakonomics seal of approval later, the choice thesis has become something of a commonplace. Today, then, the news story would not be that the proliferation of consumer choice is paralyzing us, as Schwartz argued, but that he’s wrong.
And indeed, that’s been the counterattack lately, which came to my attention the other day when economist Justin Wolfers tweeted this:
— Justin Wolfers (@JustinWolfers) January 23, 2014
(The link is to a Financial Times article that sits behind a pay-to-read firewall.) But, when I asked him, Barry Schwartz was gracious enough to respond to the pro-choice literature that’s been coming out recently, and in the process, to summarize it. We’re much obliged.
Barry Schwartz: It seems a simple matter of logic that if people have more options in a choice domain (cereals in the grocery, shirts in the department store, mutual funds in the financial market, health insurance plans under the Affordable Care Act), they’re better off. People who don’t care about added options can ignore them, and people who do care may be able to find the perfect thing. Adding options is what economists call a “Pareto improvement,” making some people better off while making nobody worse off.
Because of the “obvious” truth of the proposition that more choice makes us better off, it was big news when Sheena Iyengar published a series of studies more than a decade ago showing the opposite. Iyengar found that there are circumstances in which adding options reduces the likelihood that people will select any, whether the decision in question is trivial (gourmet jam) or very significant (401k participation).
This counterintuitive result attracted a great deal of attention, and I wrote a whole book about it, “The Paradox of Choice,” which attracted even more. (Indeed Paul Solman did a lovely piece with me about the book for The News Hour.)
Research on the phenomenon continued, extending its scope, but also identifying its limits (e.g., for people who know a domain well, more choice seems better than less, and if options are organized into categories, the too-much-choice effect is mitigated or eliminated.)
But then Benjamin Scheibehenne and two colleagues published a thorough analysis of all the existing studies and concluded that the original findings were not very robust. Indeed, averaged across all the studies they could find, the average effect of choice set size was close to zero.
Several writers have jumped on this result as evidence that the initial findings were just more “junk” social science that doesn’t replicate, suggesting economists and the rest of us should rest easy with their assumption that more choice is always better than less. A few months ago, Derek Thompson published an article in The Atlantic titled “More Is More: Why the Paradox of Choice Might Be a Myth.” Referring to what has become the classic piece of research on this topic, by Iyengar and Mark Lepper, Thompson wrote this:
It could be one of the most memorable economic studies of the last half century. Researchers presented an array of tasty jams and enticed shoppers to buy a jar. In one version, there were six varieties shown to shoppers. In another, there were 24 jams. The second, larger array attracted more traffic. But the smaller array led to ten times more purchases. Sometimes, they concluded, too many options repel us. The researchers called it ‘the paradox of choice.’ You might call it ‘feeling overwhelmed by options.’ But some economists are calling it something else: ‘complete hogwash.’
Thompson then described the influential paper by Scheibehenne and colleagues and discussed a recent paper by Daniel Mochon that showed that people hate the absence of choice; they have a “single option aversion.”
In November 2009, Tim Harford published a similar piece in the Financial Times. He did not call the original findings “hogwash.” Instead, he said that “offering lots of extra choices seems to make no important difference either way.” And he appealed to an argument that I often hear from economists: if the too-much-choice effect were true, we’d see marketers trying to take advantage of it by simplifying their offerings. Yet, a glance around the grocery store suggests choice proliferation seems to be continuing unabated. So it can’t be true.
(This is an argument reminiscent of the old joke about the economist who says the piece of paper lying on the ground that looks exactly like a $10 bill can’t be one, because if it were, someone would have already picked it up.) Harford further suggests that there is just no evidence in the real world that reducing options increases take-up.
So is the too-much-choice effect “complete hogwash” and pseudoscience?
Scheibehenne, the main source of doubt about the effect, doesn’t think so. As his paper makes clear, though the average effect size is tiny, this average is made up of many studies that show large effects — in opposite directions. That is, sometimes, more choice is better, sometimes it’s worse. And Scheibehenne’s effort to figure out when you get which effect left him, and the rest of us, without a clear answer.
And there is evidence in the real world that reducing options increases sales. After my book was published, I gave lots of talks to various industry groups and heard two striking examples. First, a large retailer of office supplies reduced the number of options offered in its print catalog in many product categories. It did this not because of the research on too much choice, but to save money on production and postage. It assumed that the change would lead to reduced sales, but hoped that production and distribution savings would outpace sales losses. What the company found was that in virtually every category in which options had been reduced, sales increased.
Second, a very large home builder, with semi-autonomous branches in many different parts of the U.S., reduced the number of options available to home buyers after they had selected a model and went about customizing it. The way the company operated, home buyers would customize their homes, advised by a consultant, in a design center. Home buyers faced 24 backsplashes for kitchen counters, 34 tile floors, 17 ovens, 21 refrigerators, 9 master bath tub packages, 13 master bath counters, 159 carpets, 37 hardwood floors, 41 vinyl sidings, 150 kitchen cabinet styles, 65 countertops, 21 kitchen faucets, 43 bathroom faucets and 26 fireplace options, among other choices.
On average, consultants spent 20 hours with each home buyer, outfitting the home. The company dramatically reduced options in many of these categories, again as a cost-cutting measure. The results were striking: reduced paralysis (four hours with a consultant rather than 20), more upgrades, less regret and more customer satisfaction. The streamlining also enabled the home builder to build homes more efficiently and economically because the construction crews could work faster with fewer errors when there were fewer variants available. These cost savings were passed on to customers in the form of no-cost upgrades; that is, the “standard” models contained features that would have been priced upgrades before.
In both of these examples, choice reduction was undertaken to save money, not to increase sales. But increase sales they did, and in the case of the home builder, they increased customer satisfaction as well.
And Iyengar, the author of the original jam study, has published evidence of a similar result when it comes to employee participation in retirement plans. When there are lots of mutual fund options available, fewer people participate than when there are only a few, even though by failing to participate, employees pass up matching money from their employers.
Similar results have also been found by Tim Rice and Yaniv Hanoch in studies of sign-ups for the Medicare Part D prescription drug plan. I can tell you that based on this research, if I were designing the Affordable Care Act and hoping for large enrollments, I would certainly have offered people fewer options than are available in most states.
So too-much-choice happens. It just doesn’t happen all the time. And we don’t yet know when it does and when it doesn’t. What should we make of this unsettling uncertainty? Do mixed results like these discredit the science that produced them?
Let me return to Thompson’s piece in The Atlantic. He says that “it’s widely assumed that overwhelming people with options — whether in TVs or delicious jams — can make them less likely to make a decision.” What is striking about this sentence? Well, prior to Iyengar’s pathbreaking jam study, a mere 13 years ago, not only was this assumption about choice overload not “widely shared,” it was non-existent.
If you asked any economist about the possibility, you’d hear something like, “Nonsense. Adding options has to make people better off. If you don’t care about all the options, you’ll just ignore them, so that for you, it’s neutral. But for someone dissatisfied with the common options, adding more will be an improvement. So some people benefit and no one suffers. QED!”
Thus, it’s not surprising to read Thompson’s claim that economists call the effect “complete hogwash.” From their point of view, choice overload is logically impossible. Forget the data. As a philosopher colleague of mine likes to say, with tongue firmly in cheek, “never let the facts get in the way of a good theory.”
Iyengar’s initial study and her many follow-ups made a real contribution to our understanding, such that a principle that was once invisible — indeed impossible — a decade ago has become “widely shared” by now. Does choice overload always occur? Of course not. Does it affect all people, in all domains of decision making? Of course not. Does it matter how options are organized and arrayed? By all means, yes. Does adding options improve decision making by making salient features of alternatives that might otherwise be ignored? Sometimes, yes. But sometimes it has a perverse effect, by making salient features of options that ought to be ignored.
In the studies of senior citizens making Medicare Part D prescription drug plan choices, it was shown that when there are a large number of plans from which to choose, decision-making quality suffers. Often people choose on the basis of essentially irrelevant features of plans, just because the relevant features are too complex to evaluate. Has anyone ever suggested that the sensible alternative to too many options is a single option? Absolutely and unequivocally not. Psychology has known about “single option aversion” for a half century. With too few options, there is the risk that none will be satisfactory, whereas with too many, there is the risk of paralysis, confusion and dissatisfaction.
The trick is to find the middle ground — the “sweet spot” — that enables people to benefit from variety and not be paralyzed by it. Choice is good, but there can be too much of a good thing. Adam Grant and I recently published a paper suggesting that this “too much of a good thing” phenomenon is pervasive in psychology.
The well-publicized failure to replicate the jam study reliably is certainly problematic. What makes it problematic is that we don’t yet know what factors determine when choice overload will occur and when it won’t. But this is in the nature of science. You could think of the history of scientific progress as just one damn mistake after another. Scientific claims are almost always wrong, principally because they are overly generalized and inadequately qualified. This is true in physics, it’s true in medicine and of course, it’s true in all the social sciences. Fallibility does not make science “pseudoscience”; it’s the nature of the beast.
The pity is that the public is so badly educated about how science works that it takes every correction or revision as a condemnation of what has come before. Lord knows there is plenty of pseudoscience out there. But with very rare exceptions, it is not produced by scientists. Someone who has a child who matured into a wonderful adult writes a guide to successful parenting. Someone who lost 20 pounds writes a book about successful dieting. These are the books that fly off the shelves. The ones describing the progress of real science, moving slowly and imperfectly toward understanding, typically don’t.
It is no doubt true that scientists sometimes seek popular audiences prematurely — before their claims have been adequately tested by peers. I, myself, may have been guilty of this when I wrote “The Paradox of Choice” a decade ago. I believe that in most cases, the reason for this is that the scientist believes she has found something out that, while hardly certain, will improve the lives of at least some people.
So, the final story on the “paradox of choice” has yet to be written. But to me, it is a beautiful example of how science works when it is doing what it should. An important idea goes from “unthinkable” to “commonly assumed.” Then, further work reveals that there are limits to this idea. Over time, we develop generalizations that are appropriately qualified and contextualized. People understand something they didn’t before, and spend their time and mental effort in ways that are more productive and satisfying than was the case before. This counts not as pseudoscience, but as scientific progress.