by Gneezy, Uri
To find out, we took a trip with our colleague Tanjim Hossain (of the University of Toronto) to the vibrant, modern city of Xiamen, in Fujian province on the southern coast of China, not too far from Hong Kong.7
Xiamen is home to lots of large factories—such as Dell and Kodak. The site of our six-month experiment was a 20,000-employee Chinese high-tech firm that produces and distributes computer electronics. The company—Wanlida Corporation—produces and distributes cell phones, digital audio and video products, GPS navigation devices, small home appliances, and so on, which are exported to more than fifty countries.
Our goal was simple: we wanted to see if we could increase productivity at the plant using simple framing manipulations. So we sent two different letters to two different groups of employees.
Imagine for a moment that you are a twenty-one-year old woman—we’ll call you Lin Li—working for Wanlida, and your job is to inspect PC motherboards. You come into the factory on Monday morning, sit down at your desk, and turn on a magnifying light like the kind dentists or surgeons might use. You pull on a pair of lightweight gloves, take a motherboard in your hands, and go over every chip, nook, and cranny, looking for defects. You do this for nine hours a day, six days a week, and, of course, receive a salary for your work.
One day, you receive a letter from management. “Dear Lin Li,” the letter says, “You will receive an RMB 80 bonus for every week in which the weekly production average of your team is above 400 units per hour.” RMB 80 is about $12, which is a pretty nice weekly bonus for a blue-collar worker in China. Because the average salary of workers in China is between RMB 290 and 375, RMB 80 represents more than 20 percent of the weekly salary of the highest paid worker. None of the 165 workers involved knew they were part of an experiment.
Feeling invigorated, Lin Li goes back to work, smiling. Another young employee—we’ll call him Zi Peng—receives a different letter: “Dear Zi Peng, You will receive a one-time bonus of RMB 320. However, for every week in which the weekly production average of your team is below 400 units per hour, the salary enhancement will be reduced by RMB 80.” Zi Peng isn’t quite sure how he feels about this arrangement, but he goes back to his desk and takes up his work with gusto.
Now, this kind of framing might remind you of the incentives that we tried on teachers and students in Chapter 4, when we said that they would lose their money if they didn’t perform well. And as you also probably noticed, this kind of framing combines a carrot (“you will receive a bonus”) with a stick (“if you don’t produce enough we’ll take your bonus away”). The message is clearly—and intentionally—mixed, because we wanted to see the effects of what social scientists call “loss aversion” at work in a real factory scenario.
When we feel we “own” something—say, social media privileges (if you are a preteen), our 1960s-era LP-album collection, our car, home, job, and yes, our bonus paycheck—the prospect of losing it makes us pretty darn unhappy.
So back at the factory, which individuals and teams performed the best? Those who, like your fictional self, Lin Li, received the carrot letter? Or those like your fictional colleague Zi Peng, who received the stick letter? Before you venture a guess, ask yourself what motivates you more: “gain-framing” or “loss-framing”? And if you work on a team with other people, knowing that the performance of each member affects the entire team’s bonus, do you work harder under the reward or the punishment framing?
Here’s what we found: just having a bonus incentive in place improved productivity. The effect was in the neighborhood of 4 percent to 9 percent for workers in groups and 5 percent to 12 percent for individual workers. These are sizable effects, considering the magnitude of our bonuses. But, more interestingly, although individual workers were not influenced significantly by the loss frame, people who were working in groups increased their productivity by some 16 percent to 25 percent above the workers in the reward framing. And, guess what? Errors and defects didn’t rise.
Overall, we found that Wanlida could effectively use simple framing to increase overall team productivity.
Would these results eventually wane over time? Would the workers slow down or stop responding to the punishment incentive? The answer was “No.” Week after week, for six months, the punishment framing increased productivity.
Clearly, the fear of loss motivated the workers more than the prospect of gain. In other words, carrots may work better if they look a bit like sticks. But who wants to work for a company that gives employees this kind of dual-handed, carrot-and-stick treatment? Well, losses are a fact of life; someone has to bear them. We believe losses are a powerful motivator. Businesses have used the threat of layoffs or firing to encourage productivity, but outside of those large-scale threats, companies rarely use loss-framing.
Of course, if you are a manager, you don’t have to use incentives as devilishly designed as the ones used in this study. Remember: it has to do with framing. If you give workers a stake in their production and then focus on the losses that could come from their lack of production, you should achieve the effects described above, without scaring employees through manipulative incentives.
So What’s the Big Problem?
So why don’t businesses experiment more? A number of barriers make implementing experimentation in firms difficult. One barrier, as Scott Cook pointed out to us, is that the people in power like to hold onto their PowerPoints, and they don’t want the little guys pointing out that the emperor has no clothes, or that he might run his empire in a different way.
Another is sheer bureaucratic inertia. For example, in the summer of 2009, we recruited some students to help us with a field experiment on incentives in a large company. The company came over to San Diego to meet with us, explained the simple problem they were facing, and agreed to run the experiment within a couple of months. Four years later, the study is still buried somewhere in the big organization, waiting for management approval.
Other times, managers are intimidated by the uncertainty involved in a change and the unknown. Going the traditional route without introducing new methods is familiar, and as long as it works, it seems safer (“if it ain’t broke, don’t fix it”). Managers also feel they’ve been hired to provide solutions and make tough decisions to enhance the firm’s performance. In other words, they feel they are expected to have ready answers for the challenges the firm faces. Opting for experimentation may appear to imply that they don’t, and could compromise the appearance of their expertise—making it look as if they have failed to do their jobs.
One could overcome these barriers in two distinct ways: top-down and bottom-up. First, the company’s managing team would need to overcome the typical “short-term earnings first” mindset and encourage (and even reward) experimentation that could improve the firm’s performance, as Cook and McCallister have done. This approach requires hiring and training people to design and run experiments, analyze the data, and draw conclusions. Under a bottom-up approach, lower-level managers could conduct smaller-scale field studies and then present the results to the management, providing them with the costs and benefits associated with running the research.
Changing a tried—if not so true—mindset is no small feat. In the end, it takes a combination of fearless leadership, training, and hands-on experience to develop an experimental culture. If companies succeed at doing that, they can reshape their industries altogether.
We have seen too many top executives fall in love with their own ideas and then unleash these on an unsuspecting world, generating a big backlash, as Netflix (and other companies before and since) did. We’ve seen business leaders apply carrots and sticks in an effort to raise productivity, to no avail. We’ve seen companies try to figure out the right price for a product, without having any idea what it’s worth to consumers. These costly mistakes occur all the time, and they are utterly preventable.
By contrast, businesses small and large that do run field experiments are making more money and attracting more custome
rs. Intuit has expanded its market by testing out small ideas, and expanding on the good ones. Humana found that by actively helping senior citizens with their prescriptions and self-care, older people could stay out of hospitals and the company could save millions of dollars in the process. A big technology company like Wanlida learned that offering employees a bonus and threatening to take it away raised productivity dramatically. A small vintner in northern California, experimenting with pricing for his wine, discovered that he had been charging half of what customers were willing to pay. And Disney learned that letting people pay what they wanted for a photo taken at the end of a ride worked especially well when half of their donations went to a charity.
The bottom line for business is this: Do you want to make more money? If yes, then run field experiments. Do you want to go down in the annals of great companies? If you do, then run field experiments.
EPILOGUE
How to Change the World . . . or at Least Get a Better Deal
Life Is a Laboratory
Nearly four hundred years ago, Galileo performed the first recorded laboratory experiment. He set heavy balls on a rolling plank and sent them speeding down to test his theory of acceleration. Since that time, laboratory experiments have been a cornerstone of the scientific method. The principle of science and the test of all knowledge, according to noted theoretical physicist Richard Feynman, is the experiment. “Experiment,” he said, “is the sole judge of scientific ‘truth.’” Increasingly, economists have turned to the experimental model of the physical sciences as a method to understand human behavior.1
To date, this pursuit of the experimental method has taken place largely within the confines of the lab. Lab experiments changed the way economists view the world, as was acknowledged by the Nobel Committee in awarding the 2002 prize to Daniel Kahneman and Vernon Smith. But the strong reliance on testing behavior exclusively in the lab is changing.
We are part of an emerging group of economists using field experiments to learn about the world. While we are waiting for our economist friends, as well as those in other academic disciplines, to take up the gauntlet we have just laid down, you don’t have to sit on your own hands. You can use our tools in everyday life to discover what really works, in everything from potty-training your toddler to running a multinational corporation.
So how do you start?
First, think about the outcome you want to change. Maybe your goal is to increase the bottom line of your business; maybe it’s cajoling your child to work harder in school. You may want to help the March of Dimes raise more money in a walkathon, or find ways to cut down on your energy costs. Having a clear-cut idea of exactly what you want to change and how to measure it is crucial. The same is true for businesses: focus in and then measure. For example, grades and test scores can be measured, as can watts of energy and productivity.
The next step is to dream up a few ways to get whatever you’re measuring to change. Generally, we start from the premise that incentives matter. Simple financial incentives are great, but nonfinancial incentives can sometimes have a bigger wallop. For example, if your third-grader is obsessed with playing video games, you might be able to use that to your advantage. Offering extra video game time in exchange for higher homework scores might be priceless in the eyes of someone so young. (This approach won’t work for all children, though. Our findings suggest that as kids get older, nonfinancial incentives have less power. But use your own experiments to see what best applies to your particular situation.)
Sometimes, removing bad incentives can make a huge difference, too. For example, if in your apartment building there is one electric meter for all residence, and you split the bill equally, then bad incentives are in play. As we noted in Chapter 1, splitting up checks equally may cause people to consume more than they otherwise would. Replacing this incentive with something more sensible (like individual meters) can do a lot to reduce needless expenditures, to say nothing of bad will.
Once you have a plan in place, all you have to do is apply a bit of coin flipping, or randomization. You’ll want to compare the difference in results between a “control” and an “experimental” situation. For example, if you devise two strategies for negotiating a lower price with used car dealers, flip a coin before you go to each dealer to choose what approach to use with which dealer. Flip “heads,” and you make the first offer to the car dealer. Flip “tails,” and the dealer makes the first offer. When do you get better deals? If you want to learn even more, go to a third dealer and make your first offer to him. Or try this: tell some dealers, “I will be visiting five dealerships today.” Let others know that “this is the only dealership I’m visiting.” Then see what happens.
Alternatively, let’s say you like shopping for antiques, and you visit a few places where you can negotiate prices. In one case, let the sellers know that you do not have time to haggle, and you need their lowest price on that nice little 1790s-era dresser, for example. In another shop, let the haggling process proceed naturally. In which case do you get the lower offer?
Or let’s say you want to increase donations to the nonprofit for which you volunteer, and you’re helping out with a direct-mail campaign. Try randomly sending half the prospective donors on your mailing list a notice of a matching grant.2 In all these cases, randomization is key; the object of the game is to rule out competing hypothesizes that might influence the results of your experiment.
One of the nicest things about running an economics experiment is that you don’t need a Ph.D. to put yourself in the shoes of someone who is participating in your study. Let’s say you are on a business trip, and you leave your hotel room for the maid to clean up the day after you arrive. The first day, don’t leave a tip, and then do a quick inventory to see how neat the room is when you return to it. The second day, leave a few dollars, and then see whether your room is any tidier when you return than it was the first time. On the third day, leave a larger tip, and so on. You may find that you wind up with a few extra chocolates under your pillow on the third day. This experiment may help you decide how to tip on future trips.
Alternatively, try this experiment when you are hosting a dinner party. Instead of serving wine from bottles, serve different-priced wines from various decanters, and then ask your guests to pick the wine that they think is the best. This experiment is a great way to discover which wines to serve next time—and you may well discover that the less expensive wines are the ones that you and your guests like the most.
As you can see, we believe that the tools of economics can go a long way toward solving important problems in practical ways. When researchers, armed with the methods that we have presented here, get out from behind their keyboards and go into the streets, they discover things that turn previous theories and assumptions on their heads.
Instead of practicing a dismal science, economists can discover that they are practicing a passionate one—one fueled by deep personal interests, dealing with the human emotions, and able to produce results that can change the world for the better. But the opportunity for change goes well beyond economics. We believe that there are massive opportunities for researchers in sociology, anthropology, business, education, and many more fields to use the tools of field experiments in economics in ways that can substantially alter the lives of millions around the world.
Over and over in this book, we have suggested that the reasons why we, as a society, have not made significant progress in battling big, stubborn problems in education, discrimination, poverty, health, gender equity, and the environment, among other areas, is that we have not made a truly concerted effort to leave assumptions behind. We have failed to seek out and discover what works and why. We keep missing the opportunity to bring the tools of scientific research to understand our most pressing problems. Without understanding that life really is a laboratory, and that we must all learn from our discoveries, we cannot hope to make headway in crucial areas.
But to paraphrase John Lennon, we’d like you
to imagine an alternative. Imagine what would happen if thousands of researchers all over the world applied the same scientific methods we’ve described in the preceding pages to big problems. Imagine hundreds of experiments running in tandem all over the world, all dedicated to overturning the rocks and probing inside the closets of the greatest problems we are facing. Imagine what could happen if, once huge amounts of feedback had been collected, we could test, and test, and test again to discover what really works and why. And imagine what would happen if, armed with this knowledge, governments around the world could make broad policy changes based on this solid empirical tests.
So there you have it. You get the idea. Experiment! Go out—white coat and pen-protector are not required—and find out what’s really going on. And then let us know what you discover, or how you’ve begun to think differently.
ACKNOWLEDGMENTS
Conducting field research requires long hours, many spent away from home. The content of this book has been gathered over many years and in many different parts of the world. This book would not have been possible without the support and encouragement of our wives, Ayelet and Jennifer. Words cannot express our gratitude to them.
We also have to come up with theories, make sense of the data, and write long academic papers. For leaving us alone during those long hours at the computer, we would like to thank our children.
It has truly taken a village to produce this book. Although they are too numerous to name here, we sincerely thank our many coauthors, research assistants, and colleagues for allowing us to pursue our dreams. Without your help, none of this would have been possible. For giving us our start in the Academy, we thank our advisers Eric van Damme and Shelby Gerking.