Join us in Seattle on June 26-29 for the NCEE Leaders Retreat. Learn more here.

By Marc Tucker

An interview with Dylan Wiliam, Emeritus Professor of Educational Assessment at the Institute of Education, University of London, on his paper entitled Optimizing Talent: Closing Educational and Social Mobility Gap Worldwide, published last year at the Salzburg Global Seminar in Austria.

DylanWiliamMarc Tucker:  In your paper, you start out making an argument that today’s children are more intelligent than their parents and their grandparents and you combine that with an argument that the quality of teaching in government-funded schools appears to be higher than that in private schools in most wealthy countries.  Can you tell us more about the research on both points?

Dylan Wiliam:  The first argument draws on the work of psychologist James Flynn (the Flynn effect), an American living and working in New Zealand.  He found that IQ tests need to be re-benchmarked every decade, because IQs are rising, about 3 to 4 points every ten years.  So IQ norms are rising, and people are getting smarter in ways we may not entirely realize.  The average would be around 110 or 115 if we didn’t adjust it.  It has risen 15 points since World War II.  This is occurring on some tests more than others; arithmetic scores have gone up very little while spatial scores and problem-solving scores are increasing substantially.  Maybe young people aren’t using their intelligence today as well as they could be but there is evidence that they are smarter.

Tucker:  Most American teachers think about intelligence in the way they were taught to – it is a function of the genes.  Is the gene pool changing, or do we have a different idea now about what these tests are measuring?

Wiliam:  Research in genetics shows that the nature vs. nurture debate is essentially an irrelevant question.  Interestingly, what you find in that debate is the estimate depends on the variation in environment.  It is an unstable question unless you can calibrate the differences in environment, which no one can do.  Secondly it is a question of epigenetics—the study of changes in gene activity that do not involve alterations to the genetic code but still get passed down to at least one successive generation.  What we are seeing is that certain genes get switched on or switched off depending on the experience we have.  For example, there is one gene that if you have two long forms, nothing happens but if you have two short forms the likelihood of getting depression increases in an abusive environment but the likelihood decreases in a healthy environment.  So what we understand now is that the nature vs. nurture debate is completely irrelevant because different environments switch on or switch off different genes.

The plasticity of IQ is much greater than was previous imagined.  There were some experiments with orphans from Romania and they had extraordinary cognitive deficits when they were in the orphanages, but rich environments helped them to catch up very quickly.  Supportive environments really do help make-up for a lot early on.

ToddlersEnvironments vary quite a lot.  Many people are familiar with the Hart and Risley study, Meaningful Differences: An average three-year old from a professional-class family would have accumulated experience with almost 45 million words while an average three-year old from a welfare family would have accumulated experience with about 13 million words, resulting in a 30 million word gap.  A child from a working class house that is in the top ten percent of cognitive development will be overtaken by a child from a middle class house that is in the bottom ten percent of cognitive development by age six. So there are extraordinary differences in the environment that make any kind of speculation about genetics questionable.

I have a goal that we should magnify the impact of genetic effects on IQ because if we give all students a rich environment, then the only difference would be in genetics.  The important thing is that high quality environments do seem to make a big difference.  There are debates about things like Head Start—IQs go up while in the program and decline when they leave that environment, but if they learn to read while in the program, those skills are there for life.

There is the argument that what you do with the talents you have is more important than the actual talents you have.  Teaching kids about delayed gratification, persistence, how to apply themselves— these things are important and we are finding they can be learned.  My conclusion is that there is almost certainly a genetic component to IQ, but we can’t change it so we shouldn’t worry about it and it’s probably not that important.

Tucker:  Your second proposition is that when you take everything as equal, public schools do better than private. Can you explain that in more detail?

Wiliam:  The 2006 OECD data shows that in most countries, kids in private schools outperformed kids in state schools.  In some countries, the gaps are quite large.  But the most extraordinary thing was they had a lot of background about the educational background of the students and have complied an index of socioeconomic deprivation.  When you control for the socioeconomic status of the student and the schools, there is not a single country in PISA where kids in private schools outperform kids in public schools because you are removing group effects.  The effects of the private schools outcomes are peer effects.  So if you’re a parent and can afford to send your kid to a private school you probably should because the teacher may not be good but the peer group is.

Once you control for these things, school effects are small.  By the time kids are 8-years old, one year’s progress is about .4 of a standard deviation.  So the average achievement made by one student in a cohort is very small compared to the overall spread of achievement in a cohort.  The range of achievement within a cohort is ten times the average progress made by a cohort a year.

The consequence is that the differences between students are typically much larger than people imagine, and it’s hardly surprising that any differences in school effects gets swamped by this.  And the second thing is that teacher quality is one of the most important variables in the system, and if teachers are randomly distributed through the system, it diminishes school effects.  For all these reasons, school effects are quite small.  That explains why reform efforts based on changing the kinds of schools available to students are ineffective, because even if the schools are good, they are not making that much of an effect.  That is because teacher quality appears to be randomly distributed across the system.

Tucker:  In the United States, where teachers have more choice about where they teach than in other countries, what you see is teachers with more seniority and experience choosing the higher status and easier positions within a school, and teachers with better reputations preferring to teach in a school with more advantaged students.  So you would expect to see better teachers teaching in schools with more advantaged students – a systematic bias toward having good teachers in more advantaged schools and bad teachers in less advantaged schools.

Wiliam:  It might be true, but it might also be the other way around.  The fact is, those teachers with seniority may not be any better than the others.  Teachers with seniority may be able to migrate to easier to staff schools, but they aren’t likely to be any better – those decisions are made on things only weakly related to teacher quality, like experience.  So it doesn’t distort the system.

Tucker:  Your logic line begins by saying, in effect, that kids’ intelligence is steadily improving and we have every reason to believe that public schools are at least as good as private ones, so you ask, why are employers so unhappy?  And the answer is because the dynamics of the global economy are changing their requirements.

Wiliam:  People haven’t really understood how rapidly the world of work is changing, because it has happened incrementally.  In the 1980s, being able to type in bold on a word processer would increase a secretary’s salary by 25 percent, now, we expect 7-year olds to be able to do that.  What we see is an extraordinary increase in the types of skills that people are expected to have.  More jobs are being automated, so the number of jobs that can be done without basic literacy and numeracy skills has decreased.

People forget how much more skilled people are today then they were 25 or 30 years ago, let alone 50 years ago.  There is an extraordinary destruction of jobs by automation.  Before you were basically renting your physical strength to the employer.  A factory may still be the world’s largest manufacturer but it employs way less people.  What are left are the jobs that not easily automated or off-shored. There are quite a lot of manual jobs that will never be off-shored—Hairdressing and taxi driving will always be required locally.  Middle jobs such as appraising someone’s eligibility for a mortgage – that used to be a skilled job.  Now, computers can do that more reliably, cheaper and quicker.

A Toyota automaker employee works on an engine at the Toyota engine assembly line in HuntsvilleTucker:  I always use the example of sail making, it used to be a skilled job, but now there are algorithms that will calculate every single panel in a sail as well as the measurements of the entire sail and it will tell you the conditions you can use that sail and when it will break. And it will also cut and sew the sail automatically.  As long as the work is routine, it’s automatable.

Wiliam:  Routine cognitive jobs turned out to be easy to automate.  And they are often easier to automate than routine manual jobs because computers are simpler than robots.  Shelf stocking is still done by human beings because they can still do it cheaper than a robot.  In the auto industry, there is a woman who does a job for $25,000 a year, whereas a robot arm can do the job for $100,000 a year.  As soon as the robot arm is cheaper than the worker, she will no longer have a job.  This is the race between education and technology, as described by Claudia Goldin from Harvard.  The world of work is destroying jobs faster than we can up-skill.  We have been walking up the down escalator in the past and have been able to make progress but now the escalator is speeding up and we may fall behind.  We need to walk faster and improve our schools faster in order to progress.

America is wealthy enough to give everyone in the country a very high standard of living by redistributing the current wealth.  This will not happen.  If you are a teacher in school today you should be preparing your student for a world where the redistribution doesn’t take place as well as if it did take place—in other words, we have to prepare them for the world we will think will unfold as well as the one we hope will unfold.

Tucker:  At the end of your essay, you make the point that the job of schools used to be to identify talent and let it move to the top.  Now, schools have to be talent incubators or talent factories – we can’t just identify it, we have to create it.  What does that mean in terms of what schools look like?  How do educators have to redefine the task?  What does this change look like?

Wiliam: The talent refinery model held that some kids can learn, and others can’t, and you have to figure out the ones that you should invest time in.  In contrast, the talent factory model holds that every kid has to achieve at a high level.  And many people say that that’s an impossible goal.  I think more good things will happen if we assume that’s achievable than if we assume it isn’t achievable.  I’m not saying there aren’t differences between students – there are huge differences.  So we need a school that is designed to minimize the impact of those differences, rather than to maximize them.  Giving them more time, bringing them in for weekend tutoring – the idea that the school will do whatever it takes to make sure that every child has a reasonable shot at getting reasonable proficiency in the desired subjects.  In high school, we have that model already in athletics.  A high school football coach doesn’t just cancel the season if they only have six good players; they take the students they have and make them the best football players they can be.  We need to translate that into the academic equivalent.