Why the right kindergarten teacher is worth about $320,000
Raj Chetty - Freakonomics Podcast #87
It’s probably already been a decade since everyone started talking about the importance of data. Every now and then we give it a cool new name and try to sell it as an industrial revolution. Now, with artificial intelligence entering the mix, the expectations couldn’t be higher. However, all the data in the world is not helpful if you draw the wrong conclusions from it.
This is where Raj Chetty enters the play. He is a Harvard economist trying to understand things like inequality, social mobility, extremism, and so on. His motivation? Providing children from disadvantaged backgrounds with better chances of succeeding in life. Chetty’s research combines empirical evidence and economic theory, to help design more effective government policies.
There is one particular subject echoing throughout the show that I want to share with you. But first, do you remember Goodheart’s law? If not, no problem - I’ve written about it recently here, but in short:
When a measure becomes a target, it ceases to be a good measure.
In other words: Be careful what you measure and incentivize, because it might lead to an undesired and even hurtful outcome.
This is the point where our new story ties nicely together with Goodheart’s law. When dealing with a lot of data, we are facing a similar but different problem. What if we draw a wrong conclusion from a correlation in our data set? Examples of where we confused correlation with causation in the past are plentiful, as this TED-talk is highlighting:
The mere correlation between two variables does not imply causation.
Big data is increasingly being used and hailed for its invaluable contributions to areas such as health care, entertainment, and education. But how could education benefit from data analysis and the like? Well, one influential variable of education is the type of teacher. So here is a possible way forward:
Let’s say, we look at how kindergarten teachers influence test scores later in school. Then we find either personality traits or behaviors common in these teachers who are responsible for better grades.
Halfway there, now all we have to do is hire more teachers with similar personality traits and incentivize the kind of teaching that yields better grades. Perfect! Once again - science and technology stepped in to save the day.
Well, not so fast! Thank god for critical minds like Raj Chetty. He actually did look at how teachers influence test scores and, indeed, there are some strong correlations between the personalities of teachers and their influence on students’ grades later in school.
However, when they looked beyond school, there was no correlation between the kid’s test scores and their income, health, or quality of life. But, there was however a strong correlation between all of these factors and classroom quality in kindergarten. Here is a citation from the publication:
We were surprised, then, to find a strong relationship re-emerge between kindergarten classroom quality and adult wage earnings! Even though the effect of better classes on student standardized test scores quickly faded, being assigned to a higher-quality classroom was an important predictor of students’ earnings. Remarkably, we also find substantial improvements on virtually every other measure of success in adulthood that we examined. Students who were randomly assigned to higher-quality kindergarten classrooms were more likely to attend college and attended higher-ranked colleges. They were also more likely to own a house, be saving for retirement, and live in a better neighborhood.
Wait! So test scores don’t matter and don’t predict anything later in life? I’m shocked…!
But on a more serious note, what is the strongest factor of classroom quality? You guessed it, the teacher. Let’s say for ease of argument that there are three kinds of teachers:
Teachers who are simply there, doing the bare minimum. 1
Good teachers, who care about you and your test scores.
Great teachers, who inspire you and nurture your natural curiosity. 2
Back to Goodheart’s law- If we would optimize teachers simply for test scores, we would miss out on some enormous economic value (not to mention the quality of life and happiness of millions of people). So to make a long story short, correlation is not causation and when it comes to complex systems, you need to be extra careful about what you measure and what you optimize for.
With that, I’m leaving you for now.
Have a successful day
GNF
Bonus: Read the full paper here.
It wouldn’t be an economics paper if the researchers didn’t quantify their findings. So, here it goes:
To quantify the size of these effects, we isolate the part of the class quality that is driven by teachers. We estimate that going from a below-average (25th percentile) teacher to an above-average (75th percentile) teacher raises a child’s earnings by about 3.5% per year. In present value, that adds up to more than $10,000 in additional lifetime income on average for each student. When you multiply that by 20 students in each class, the additional lifetime benefits from a single year of high-quality kindergarten teaching is about $320,000. These are huge stakes at play and underline the importance to the nation of having high-quality classrooms and schools. The benefits of classroom quality for adult outcomes is not limited to only the kindergarten year. High-quality classrooms in grades 1, 2, or 3 had a similar beneficial impact. We do not have the data to allow us to determine whether classes in grades after 3rd grade have the same effect, nor can we say anything in this study about preschool education. But we think our results point to the importance of the early grades in general and not about kindergarten in particular
And who can blame them? Have you ever been the only adult in a room with more than five kids? It’s a warzone! - Duck and hide! I don’t know why people become teachers in the first place, but God bless them.
If you have watched “Dead Poets Society”, I’m guessing you are thinking of Robin Williams as the teacher.