Performance-based pay took another hit this week. A study of the much vaunted New York City system by Harvard economist Roland Fryer found that teachers of many kids in the program actually did worse over time.

This certainly isn’t the first public failure for merit pay. But it’s an important failure for three reasons. For one, most people seem to take it for granted that merit pay schemes will work as long we try enough of them on enough teachers for enough time. And yet, the last few studies that have come out have been negative or inconclusive. The fact that New York City Mayor Michael Bloomberg was hugely enthusiastic about the program when it was launched – and had now discontinued it – makes you wonder how a pretty smart guy when it comes to money could he have been so wrong. And the program wasted philanthropic and tax dollars on an idea that has little research support.

By most standards, this was a fairly large study jam-packed with 20,000 teachers and their students, more than enough to get a pretty good take on the notion that more money might encourage better teaching. But perhaps most significant of all is that Fryer, who was predisposed to supporting merit pay, came out of it with both his integrity and an “ambivalent” view of performance pay.

So here we are, 10 years into “scientifically-driven research-based” reform and we’re still pushing programs that, when researched, don’t seem to work. This is patently inconsistent with how we say we are fixing our schools. We’re still misusing public funds and abusing the public trust. It is obvious that so-called “research-based” programs and practices either aren’t based in good research or have never been proven effective at all. A line is being drawn down the center of education reform, a battle line. And it isn’t about right or wrong. It’s about ideology. In this case, Bloomberg put it perfectly when told that “I am a capitalist, and I am in favor of incentives for individual people.”

Still the question persists: Why doesn’t merit pay work? Here in America, capitalism is an ideological slam dunk. Why doesn’t paying teachers more money lead to better performance? I’ve surveyed dozens of groups of teachers informally about this during my career as a teacher trainer. I’ve also observed, as I’ve had to motivate teachers myself, what works and what doesn’t. I’ve made far more mistakes than Bill Gates or Mayor Bloomberg. I just haven’t made them with millions of dollars or thousands of people.

Even in my own company, where I’ve hired teachers to be consulting associates, money has never been the mechanism for improving performance. And as a former technology entrepreneur, I have been through the same intellectual frustrations about why this isn’t so. At the same time, I’ve also figured out why money doesn’t motivate teachers and what does.

I’m not saying money will never work or that money never has some positive effect. But my experience of teachers and teaching—and merit pay programs—tells me that this lever of reform will never deliver all the leverage we think it should.

Over the years, I’ve identified many reasons why I don’t think merit pay will be very effective in improving teacher performance. But here are a few simple ones that I think most people will see some merit in—even if they’ve always considered themselves ardent supporters of merit pay:

In general, teachers are not an economically-sensitive group. Teachers actually make a much more reasonable living than most people think (especially when we look historically at job security and benefits) when compared to other professions requiring similar knowledge and skills. They usually make enough money to live reasonably well at the lower levels of the middle class, especially if they have a working spouse, they teach more than 10 years, and they get a Master’s degree.

But those who choose to be teachers are not highly motivated by monetary gain. One of the things teachers like best about teaching is that they can know exactly how much money they’ll get on the first day of every month. They usually don’t have much experience negotiating their own pay. They will take whatever it is offered and will not be especially concerned with what that is. They are not MBAs using complex calculations to squeeze out increased profits.

For people who are not highly focused on increasing wealth, and who tend to think they are always poorer than they probably are, fluctuations in income, even if there’s a possibility of increasing it, are difficult to deal with and best avoided. When income shifts, even in a positive direction, budgeting and financial discipline come into play; it’s possible to plan badly and make a mistake. Most teachers feel more comfortable knowing exactly how much they’re going to make every month—and neither modest increases nor even modest decreases are as important as consistency, predictability, and the minimization of risk.

Teaching is a complex act that is difficult to define and to track in process. Studies, most notably a large one done by The Federal Reserve, show that performance-based pay works reasonably well for simple tasks and for piece-work tasks. For complex tasks, however, it seems to have little or no positive effect at all.

As teachers perform what is an unpredictable, ill-defined, and hard-to-track task, all of their energy goes into managing its complexity. It’s a lot easier, in a simple task, especially a piece-work task, to have cognitive bandwidth left over for keeping track of how well you’re doing. This “keeping track” is what provides the actual instances of motivation that, over time, inspire better results.

As a result of teaching’s complexity, it’s hard to know how you’re doing relative to a quantitative goal from week to week or month to month. Are the kids proceeding quickly enough? It’s hard to know for sure. As a result, you don’t get the rush of motivation from knowing that you’re ahead of schedule. Nor do you get the information you need to motivate yourself to catch up if you realize you’ve fallen behind.

Many teachers do not value or trust test scores as legitimate measures of learning. Teachers value learning, even when they’re not sure what learning is. But what most believe in their heart of hearts is that tests don’t measure it. They may be wrong about that psychometrically. Then again, most teachers aren’t psychometricians. So when merit pay plans are based in whole or in part on improving test performance, teachers are less motivated than we might hope they would be simply because they don’t trust or value the metric they’re being measured by.This occurs in the private sector all the time when thoughtless managers create incentive programs that don’t match the value systems of their employees.

Many teachers don’t trust tests. And it’s hard to imagine they ever will simply because testing is so brutally reductive. Turning kids’ learning into numbers removes so much of what teachers feel they pour into their jobs, what marketing genius Seth Godin calls “emotional labor”. You can’t be even an average teacher and just phone it in every day. The kids expect you to care about them, even if they don’t always seem to care about you. Teaching is a deeply emotional endeavor.

Standardized tests are particularly effective at erasing our ability to account for the emotional labor component of a teacher’s work—or the emotional competencies of a teacher’s students. But these may well be the very components many teachers value most. In an age of data, we may be able to measure what teachers do. To psychometricians, it may make perfect sense to reduce kids to dots on a graph. But for many teachers, it will never make sense.

Most merit pay programs are insufficiently transparent. In most performance-based pay situations, the performance-pay mechanism is transparent. You can see it all the time and you can use your own frequent assessments to motivate yourself. This idea of transparency lies at the heart of what makes incentive systems work. If I sell cars on commission, I can calculate my earnings to the penny any time I want. If I’m a running back pushing for my 1000-yard bonus, I can see my totals at the end of each game and know exactly what I need to accomplish during the rest of the season. If I’m a CEO with my pay tied to profits or stock price, I can watch those numbers all the time, too.

In performance-based pay programs in education, however, teachers often don’t know if they’ve been successful until a school year is almost over. At that point, no adjustments can be made. You don’t know where you are, if what you’re doing is working, or how exactly to change things if what you’re doing isn’t working. To most teachers, it seems as though they have very little control over whether they receive your bonus—and no way to use their own assessments along the way as potential motivators.

Most human beings do poorly in situations where they feel they have little control over their results. This may account for why, in some performance-based studies like the recent NYC affair, some teachers actually perform worse. It’s also why so many teachers do not support merit pay at all. They just can’t tell at any time how much money they’re supposed to be making and, not being “money people”, then tend to think they’re getting taken advantage of if they can’t be sure they aren’t.

Most teachers are already doing about as well as they can. I’m over six feet tall and I love to play basketball. But you could offer me a million dollar bonus for a single slam dunk in a low-skill rec league and I’d never collect a penny. Most serious players over six feet tall can dunk. But I’ve just never been a good jumper. And regardless of all the things I did in my competitive years to jump higher, the gains I achieved never even got me within a few inches of touching the rim.

Most teachers are in a similar position. They’re maxed out. It’s not that they couldn’t learn to be better. But it isn’t money that’s going to do the teaching. It’s probably training, support, better working conditions, a docile and good-natured group of kids, a great principal, good colleagues, collaboration, a lot of self-study, a lot of reflection, and a lot of time—all things that rarely come in performance-based pay programs. They are unlikely to ever earn a bonus.

There are two important implications here. First, we know that teachers are poorly trained and poorly supported in their schools. Second, and more important, is that teachers know this, too. I don’t think I’ve run across a group of trained college graduates with a deeper sense of shame about their abilities, a greater sense of resentment about the system in which they work, or more powerful and all-consuming degrees of cynicism, pessimism, and victimhood. The idea that somehow teachers don’t know how ineffective and unmotivated they are is naive. As a teacher, I know, most of the time, whether kids are “getting it”.So, no matter how much you incentivize me with money, I already know I’m probably not going to be able to win the bonus because I’m already getting the best results I can imagine.

Most teachers do their best most of the time. Whether their best is good enough is not the question at hand. The question is how to motivate them to do better, and even to exceed their own highest expectations. Performance-based pay has little effect on this because there is no significant surplus of unused teaching capacity. You can’t get much higher than 100 percent.

We have to help teachers become better at their work and being able to do better without exhaustive effort. This is a simpler and better way to reward teachers and motivate them to improve their work than current performance pay plans. Tomorrow, I will discuss how I do this in my work training teachers.