Incentives – Standardized Testing

Would a teacher ever cheat?

Of course this question will raise a few hackles. However, teachers are human, and humans respond to incentives. We do things that reward us, and avoid things that punish us. I’m intending to write a few blog posts about how incentives affect education – for good or for bad.

So, would a teacher ever have an incentive to cheat?

Let me ask another question – suppose the government or regional educational council or whatever decided to reward good teachers. Would this ever inadvertently create an incentive for teachers to cheat?

In my previous post, I discussed the question of how to reward good teaching. I pointed out that there are problems with any attempt to do so.

  • Rewarding teachers purely according to whether their pupils are taught well is impractical. The outcome of truly good teaching is a successful adult, but this only becomes evident long after the teaching takes place – much too late to meaningfully reward that adult’s early teachers.
  • On the other hand, good teaching produces some immediate changes in the student. Unfortunately, there may be other ways to produce those changes. If you reward teachers who produce those changes, it may merely bring those “other ways” to light.

One example of the latter is standardized testing. In standardized testing, every student in an educational system is given the same test. They are tested again in later years. All test results for all students are fed back into a central computer system, so that the performance of individual students (and their classrooms, schools and districts) can be  tracked over time. An example is Australia’s NAPLAN tests, issued nationwide to students in grades 3, 5, 7 and 9. The individual results are sent home to the kids. The aggregate results for schools are (controversially) posted on a publically available website.

Now, it is true that “classrooms of students with good teachers will perform better on standardized tests”. If the test results were buried in a vault and never opened until 2085, it would also be true that “classrooms of students that perform better on standardized tests have good teachers.”

Unfortunately, as soon as an educational authority actually tries to use the test results to move money around, it opens a can of incentive-based worms. Clearly, it’s also true that “Classrooms of students with teachers who cheat will perform better on standardized tests”. If, as was attempted once in Chicago, teachers are rewarded when their students do well on standardized tests, then good teachers will be rewarded. So will teachers who cheat – especially those who cheat just a little so they don’t get caught.

With rewards based on standardized test results, it’s up to the teacher to decide which of the following is most worth the effort, risk and conscientiousness involved:

  • make an effort to teach well
  • forgo the potential reward
  • cheat

In a case study described in this book, it was found that, indeed,

  • If you reward teachers when their students do well, then some cheat.
  • If you punish teachers who cheat blatantly, then some still do, but not so many.

In the case study, the educational authority (in Chicago) found that perhaps a few hundred teachers had cheated by modifying their students’ test answers! Only a dozen or so cases were blatant enough to take action, and so a handful of cheating teachers were fired. The next year, cheating dropped by 30%. The potential reward for cheating had dropped relative to the risk involved, changing the relative incentives of the three options each teacher faced.

However, even without cheating, there are ways for schools to manipulate the test results. Even in my son’s school, which is an excellent school, they spend significant amounts of time getting the kids to work on practice tests. They could, instead, be using that time to shore up other (non-tested) skills that might be lacking in the kids. No doubt this kind of thing is happening almost everywhere. In fact, this is one of the reasons schools are opposed to the national testing system and the way the results are being used.

The occasional school that says “we won’t ‘teach to the test’, we’ll save that time and give the kids the education they really need” will find that their test results don’t reflect the fact that they are actually getting a better education than the very similar school down the road.

We should expect, therefore, that

  • whenever a standardized test instrument is introduced to gauge teaching quality, and
  • the results of that test come to affect schools or teachers,

the test instrument immediately becomes a less-than-perfect gauge, no matter how accurate it seemed before it was introduced.

The only solution I can think of for policymakers who want such a gauge is

  • chop and change the test each year – make it a surprise until the day it’s given to students. That way, schools and teachers can’t anticipate the questions and ‘teach to the test.’ Unfortunately, it also makes it hard to compare kids’ performance from year to year.
  • don’t let schools and teachers administer the test – so there is little opportunity for them to manipulate the test results in other ways. Unfortunately, this would introduce explicit administrative costs (although there’s no reason why these would bigger than the hidden administrative costs of asking teachers to administer them), therefore harder to justify politically.

It’s not that teachers can’t be trusted, of course. The whole of society trusts teachers five days a week with the education of the next generation. It’s that there’s a very few bad apples who can’t be trusted, and rewards based on standardized testing make it very hard for the rest of the barrel to ignore them and focus purely on good teaching.