by Gneezy, Uri
Students in a third group were told that if they improved on their previous test score, we would give each of them $20, but not until a month after the test. A fourth group received a $3 trophy if their scores improved. And as is always the case in our experiments, we had a control group. This group was offered no reward, though we encouraged them to try to improve their scores.
Our incentives had a huge impact. Overall scores improved between 5 and 10 percentile points on a 100-point scale, putting the students on a more even footing with wealthier suburban kids. This was an amazing improvement. Even though the students had no idea an incentive was coming until just before the exam started, they improved remarkably. It showed that an important part of the racial achievement gap was not due to knowledge or ability, but simply to the students’ motivation while taking the test.
This result highlighted the importance of understanding what motivates students: though they weren’t very interested in taking the test, their scores shot up in the face of financial incentives. (Think of what might have happened had we offered these incentives and also given them time to prepare and study.) The goal of this experiment wasn’t to design an incentive scheme to be used in other schools. What we were after was a diagnostic tool that could help us understand whether the score gap was due to differences in knowledge, or differences in effort on the test itself. The answer to this question could help us design relevant interventions to reduce the gap.
That said, the incentives worked differently for the different groups. We discovered that the older students, in particular, responded well to the money, whereas the younger ones liked the trophies we offered instead. Offering a second-, third-, or fourth-grade student a $3 trophy before the test improved her performance by 12 percentile points. These effects were large; indeed, they were similar to the impact of reducing class size by one-third or considerably improving teacher quality. This is an important point, as we discussed in Chapter 1. Incentives don’t have to come in the form of money. In some situations, and for some people, a trophy (or flowers, chocolate—you name it) can go a long way.
As we expected, giving students the rewards beforehand—and threatening to take them away if their scores didn’t improve—boosted scores much more than promising to give them the money later. In fact, the students who were promised to receive $20 a month later did not improve at all. Again, it seemed that framing something as “it’s yours to lose” worked better than “it’s yours to win, and you will win it later.” To understand this, put yourself in a student’s shoes. If you are offered the financial reward for improving, you score a lot higher if, before you even take the test, you are thinking about buying those new skateboard wheels. For young children and teens, the world is all about the present; our experiment helped us understand what really motivates them.
Obviously, all we had managed to do here was to convince students to try a little harder. But we were worried. What if incentives lost their effect on behavior over time? That is, we thought we could get students to try hard a few times, but we wondered if, eventually, the incentives would lose their impact on behavior. Alternatively, would kids only try when they were offered an incentive? Would they give up if a $20 bill wasn’t at stake?
We often hear concerns from educators, parents, and policy makers that even while financial incentives may produce short-term improvements, children could be hurt in the long run; they might stop putting in any effort without compensation.9 In fact, we found no evidence of one-time rewards hurting test scores in the future. As we expected, the one-time incentives also did not lead to lasting learning gains. The simple short-term experiment did show, however, that the children were more able than one might have previously thought, based on the standard approach to testing.
The next step was, of course, to stretch our behavioral economics interventions even further. What if students were rewarded every week for an entire semester for something like independent reading? So we launched another study that offered students in seven schools $2 (or an equivalently priced nonfinancial incentive) for every book they read over the course of a semester. We kept track of their reading by letting them log onto an online program called Accelerated Reader that has short quizzes for just about every book available to students. The quizzes aren’t difficult, but it’s hard to score well if you didn’t read the book. We decided that if a student scored 80 percent or higher on a quiz for a book, they could receive the rewards, which we gave out each week. As in the test incentive study, we compared giving students the incentive either at the beginning of the week or at the end of the week. We found that the incentives in both cases increased reading by 37 percent, but the extra reading had no impact on test scores.
Can the Same Idea Work with Teachers?
Of course, students don’t learn in a vacuum. We needed to find out whether offering incentives to teachers might work, too. After all, it’s hard to run a classroom while your students are undisciplined, indifferent, scared, hungry, or absent. And it’s harder still to work so hard and face the fact that many of your ninth graders are reading at a fourth-grade level, and that fewer than half of the kids you’re trying so hard to teach will graduate.
One of the big criticisms of public education (and other public institutions) is the scarcity of incentive-based pay. In many private-sector companies, the amount you take home in your paycheck is based on your performance. Let’s say you graduate with a B.A. in business and want to go into sales. When you get a job, you generally receive a base salary with a bonus incentive. If you perform well for a year, you may receive additional benefits or even a promotion. Other incentives are available, too: if you are part of a sales team and you, as a group, sell more than your expected quota, you get a team bonus. And if your company as a whole does well, you might receive an extra kick in the paycheck.
But if you are a public school teacher (or work for the public sector in general), few of these incentives exist. Three factors determine your pay as a teacher: your level of certification, your postgraduate degrees, and your seniority. That’s it. Stick around long enough, you will be paid more, whether you are a star or a slacker.
One of the main reasons why we do not know as much as incentive programs work is because teachers unions are loathe to adopt pay-for-performance schemes. This hit home when we originally took our idea to Ron Huberman. We told him about our ideas to incentivize teachers in Chicago Public Schools. Depending on how well they did, teachers could earn up to $8,000 on top of their salaries. But the teachers unions balked. “No, absolutely not,” they told us. “We could not imagine anything like this working.” Even when Ron Huberman stepped in to try to persuade them to allow us to run our experiments, they refused.
But we still had Tom Amadio, the maverick in Chicago Heights, in our corner. With his persuasive help, we reached an agreement with the Chicago Heights teachers’ union. Fortunately, these teachers were willing to try anything to help their students.
We offered more than 150 Chicago Heights teachers a chance to earn an extra bonus.10 In one treatment, an individual teacher could strive for the $8,000 bonus; in another one, teachers working in teams of two split the bonus (the idea being that team teaching would allow them to share lesson plans and ideas). We also applied the same gain-versus-loss-framing, carrot-and-stick motivators that we’d used with the students in the computer labs. We wrote checks for $4,000 (the average reward) to some individual teachers before the school year started, with the stipulation that they would have to repay all or some of the money if student performance didn’t improve. We also wrote checks for $4,000 each to some of the teachers who were going to work in two-person teams, with the same stipulation. (Given that the average teacher’s salary in Chicago Heights is around $64,000,11 this extra $4,000 was a substantial amount. Imagine yourself getting this extra money in September, and then having to pay it back in June. No pressure, of course.)
Our results showed that when teachers were threatened with losing the rewards they had
already received, student achievement in math jumped up about 6 percentile points, and in reading, about 2 percentile points. This type of incentive seemed to work particularly well when teachers worked in teams. Overall, their students’ achievement improved between 4 to 6 percentile points.
This result was nothing short of astonishing. To put it in perspective: if the low-income, minority students in Chicago Heights could repeat such percentage-point gains for each year of elementary school, it would be enough to close the entire gap between their achievement and that of wealthier white kids from the suburbs.
Closing the Loop: Incentivizing All Actors
Having figured out how to motivate parents, students, and teachers in isolation, we now wanted to discover what would happen if all three parties—parents, children, and teachers—worked simultaneously to improve student performance. Would their combined efforts lead to improvements in student outcomes such as higher test scores, higher graduation rates, and better job outcomes? Would this kind of cooperation blow their grades through the roof? One might think so, at least on an intuitive basis. But the empirical evidence has been scant.
To learn more, we ran another experiment in Chicago Heights. The test involved elementary school kids who were at risk of failing to meet state standards.12 We worked with 23 reading and math tutors who in turn met with 581 K-8 students in small groups for 100 days. The five treatment groups included an incentive for the tutor only, an incentive for the student only, an incentive for the parents only, an incentive for both the student and the parents, and an incentive for all three—the student, the parents, and the tutor.
We assessed the students every two months, and if they met all of our achievement standards, we paid a reward of $90 each. In another setup, students, parents, and teachers shared the incentive, receiving $30 each. When we divided the reward between students and parents, they each received $45. We paid the students who met the standards immediately upon completion of their exam.
As was the case in all our other school experiments, the students were excited to participate. But because the study was only open to kids who were failing, other students were disappointed. (We heard anecdotally that some kids were thinking of intentionally doing poorly on their own tests so that they could participate. This information was troubling, but it does say something about the power of incentives to change student behavior.)
We also wanted to provide the parents with a tool for helping their children improve, so at the end of each week, tutors created a homework assignment for students to work on with their parents. After the assessments, parents attended pizza parties we held at the school. We reviewed with each of them their child’s performance, paid those who qualified, and made sure they were aware the incentive program was continuing.
We found that when we divided the incentives into three smaller $30 rewards, the improvements were relatively small. Even though everyone was given an incentive to do more, the impact just was not there. However, the $90 payout to one person was quite effective. Interestingly, it didn’t really matter whether we rewarded $90 to the students, the tutors, or the parents; as long as one of them was rewarded at the $90 level, the incentive worked well. While it clearly takes a team of people to educate a child, we found that a higher-stakes incentive for just one actor generates the biggest bang for the buck.
And the student results? Test scores shot up between 50 and 100 percent compared to the cases in which nobody received an incentive. If these results seem radical, it’s because they are: the incentive was enough to transform the test scores of the average kid in Chicago Heights into the sorts of scores typically only seen in wealthy suburban school districts.
Our explorations into public education have taught us the power of combining field experiments with economic reasoning. We learned that kids really do respond to immediate rewards, and that the threat of taking away that reward is more powerful than paying it later, both for students and for teachers. We also learned that parent participation really helps in teaching kids not only how to read and add, but also how to appreciate noncognitive skills such as patience, and how investment now leads to higher rewards later.
On the downside, we also learned that the behavior of some kids, especially kids in high school—Kevin Muncy, for example—are harder to change. Kevin was already disengaged by the time he entered Bloom Trail High; ultimately he failed all his classes. Urail King, on the other hand, ended up doing better. Borderline kids like Urail were more easily motivated by the cash and the lottery. The improvements we saw in the high schools were promising—dozens more kids graduated than might have otherwise—but they weren’t dramatic. These high school kids just didn’t appear to be as easy to motivate as we had hoped.
These insights drive home an important point: even if we offered kids like Kevin a million dollars to solve a hard math problem, he couldn’t. Why? Because by the time kids are fourteen years old, they are already tilting one way or the other. They have already missed out on important investments that make it quite difficult to achieve high levels of competency in certain subject areas. They have already been deeply imprinted by their earlier experiences; parents have lost much of their influence and their supply of coercive tools. And if at that point they are only reading at a third-grade level, bringing them up to speed is awfully hard.
If you haven’t already learned how to focus on a lesson, solve problems on your own, and stay out of trouble by the time you are in the ninth grade, your odds of success are low. For disengaged kids like Kevin, a more serious intervention is probably more appropriate.
Think of it this way: if we asked you to solve a second-order partial linear differential equation, and told you that we would pay you a million dollars if you solved it, could you? If you haven’t been trained to solve this kind of math problem, even a million-dollar incentive will have no effect. If higher-level problem solving isn’t already achieved through years of good schooling, applying incentives so late in life won’t make a difference.
This does not mean that we give up on those kids; quite the opposite. There is a productive place for everyone in our vibrant world economy. But, clearly, we need to see what will happen if we intervene with younger children. Early childhood education has the chance to give everyone an open door to the highest levels of society.
To find out what we did, turn to the next chapter.
CHAPTER FIVE
How Can Poor Kids Catch Rich Kids in Just Months?
A Voyage to Preschool
One of the longest-running programs to address systemic poverty in the United States is Head Start, serving millions of young children since it was founded in 1965 as part of President Lyndon Johnson’s “war on poverty.” Although the original intention of Head Start may have been laudable, the program has proven far less effective in helping disadvantaged four-year-olds get a jump on cognitive and social skills than originally hoped. Several academics have by now dissected the program and have found that it has several deficiencies, chiefly because the teachers are largely undereducated, underpaid mothers, fewer than 30 percent of whom hold bachelor’s degrees.1 Another problem is related to the fact that instead of being run by the Department of Education, Head Start is run by the Department of Housing and Human Services—an agency more involved in making up for the effects of inadequate education than in improving it. We suspect that presented with the totality of the evidence, reasonable people would question whether Head Start provides kids with significant benefits.
This is awfully disappointing, especially when you consider the cost: the price tag for keeping one child in Head Start for a year is about $22,600, whereas day care only costs $9,500. As Time columnist Joe Klein has noted in criticizing Head Start, “We can no longer afford to be sloppy about dispensing cash—whether it’s subsidies for oil companies or Head Start—to programs that do not produce a return.”2 We could not agree more. The question is: What would work better?
After compiling the results of our field experime
nts discussed in the previous chapter we—along with our colleagues Steven Levitt and Harvard’s Roland Fryer—had an honest conversation with Tom Amadio and the Griffins. While we had seen considerable improvements amongst the K-12 students that we touched, we had not hit a home run yet. By the time we reached them in ninth grade, for example, we might be able to help them graduate from high school, but it was unlikely that they would go on to become successful engineers; for that type of impact, our interventions occurred too late.
One idea is to reach children at a very young age, which could potentially give them the leg up they needed early on in the education process. The best way to do this while maintaining the integrity of the scientific process, however, was to build our own experimental schools to learn about the education process—what works, when does it work best, and why.
For academics like us, the entire notion of setting up one’s own school to learn about early childhood education is like building a new research laboratory from scratch. While we concluded that this was the most appropriate way to tackle such an important problem,3 building schools for this purpose presented a new challenge for us. The very first, and perhaps most important, challenge was to come up with resources. We quickly learned that the Chicago Heights school district was hardly getting by. It had few resources to even teach its own pre-K kids, much less expand to serve both its kids and those from the surrounding communities (which was necessary to obtain the sample sizes we required).
Once again, the Griffin Foundation revealed its generosity, this time with a whopping $10 million initiative to work with young kids and their parents. So the Griffin Early Childhood Center (GECC) was born. GECC consists of two preschools in one of the poorest areas of Chicago, and it is the beating heart of one of the largest controlled field experiments in education ever conducted.