Wednesday, December 24, 2008

Economist's Notebook: Education, Part 1

This post will be the first in a series of posts in which I try and distill the frontier of economic research on education.  I have talked a lot about how, as a development economist, I think that a state like Oregon should spend heavily on education as human capital is perhaps the most important investment for future growth.  

But am I right that this translates directly to a high income country?  Is more money the solution or are there other things the state can do to improve the human capital accumulation of its children?  And, should the state focus more on K-12 or on higher education?  These are all questions that I realized I did not have good answers to, because I did not have enough knowledge of the latest evidence.  

So in a series of posts over the next few weeks, I hope to focus on some particular pieces of evidence from the economics literature.  Economists do not have all the answers, of course, and bring only one perspective, but economists are particularly good at examining data and teasing out causal links, so I don't claim these to be the last words, just starting points.  

So the first paper I want to introduce to the discussion tries to tackle the very difficult problem of assessing what matters for the quality of primary education.  The difficulty lies in the fact that since families choose to locate in areas in part because of the quality of the school district or a particular school itself, if we see variation in student achievement it is hard to tell whether it is the school that is causing the high levels of achievement or the families.  So, for example, we may see that one school spends twice as much per pupil than another school and has student achievement scores that are twice as high in the well funded school.  The question is whether it is the spending per pupil that is causing the high achievement or that the families of the kids in the better funded school are better educated, wealthier, etc.  Why this matters is that if we see a poor performing school, we might be tempted to say that increasing the per-pupil spending is the answer based on the evidence above.  But we just don't know.  

So how do you tackle this problem?  Well, Rivkin, Hanushek and Kain, in the March 2005 issue of the journal Econometrica use a unique data set from Texas that observes multiple cohorts of children as they move through grades 3 through 7.  This data set allows a very serious examination of some basic questions.  Do schools matter in the achievement of children, or is it all family and peer effects?  Do teachers matter?  Do observable things like class size, teacher education and teacher experience matter?  Finally, if these observables do matter how big are the effects (important if you want to know what to focus on)?   Though these are some pretty basic questions, answers to then have been elusive.  Why?  Well imagine you see that teachers with masters degrees seem to have better performing students than teachers with only bachelors degrees.  You might be tempted to conclude that this means that teachers with masters perform better in the classroom.  But it could be that talented students and their parents seek out such teachers (probably through seeking out well-funded schools that employ such well-qualified teachers in abundance).  Also schools themselves sort teachers and students and often match well performing students to the well-qualified teachers and therefore create a self-fulfilling prophecy.  

How does this data help overcome this problem?  Well by seeing how individual students do through repeated observations over a number of years and by seeing how different groups of students perform in the same teacher's classroom allows an econometrician to control for student heterogeneity and the non-random matching of students and teachers.  Students may have many things that make them better students that are not observable (drive, diligence, intelligence, etc.) but by looking only at individual students through the course of their studies you can effectively control for the effects of these unobservables.  Ditto the non-random matching of students (as long as the policy remains the same through the time of the data).  

So what do they find?  Well, teacher quality is incredibly important.  In fact they find that exposure to high quality instruction can largely offset the disadvantages associated with low socioeconomic background.  However, the quality of teachers is NOT related to their academic background: teachers with masters were not systematically better than those without.  Also, after the initial years in the profession, experience was not related to achievement.  These results help explain a puzzle in the education literature: it is believed that teachers are incredibly important but when we look at things like education and experience of teachers they seem unrelated to student achievement.  So it is true that they are incredibly important, but education and experience are not what characterizes a good teacher. [This poses another puzzle for those who believe in the Spence signaling model, but that is another story...] 

They also find that smaller class sizes do matter, but not as much as teacher quality.  Class size matters most in earlier grades, but other school resources have little to no measurable effect. In other words increased spending on schools seems to matter through the smaller class sizes and in teacher quality, but not in teacher credentials and in other school amenities.  

This research created a policy puzzle, if good teachers cannot be easily identified by their credentials or experience how do you figure out who the good ones are?  It appears that to a large degree, you have to learn their quality by hiring them and observing them.  This argues for considerable flexibility in schools' ability to hire and fire teachers, especially in the first few years of their employment.   Being able to effectively identify the talented teachers and retain them seems to be the first-best solution, but after that, reducing class size seems to be the best way to spend money on student achievement.  I am having a hard time locating recent data on state-by-state comparisons of average class sizes, but the data I can find from the 90s places Oregon near the bottom of all states in terms of class size.  

In the next post in this series, I will look at higher education. 

NOTE: I have updates this post with a link to the article


Doug said...

Suggestion: I'd like to follow your blog but the combination of the black background and fine font make it very hard to read.

Dave Porter said...

Two points:
(1) Malcolm Gladwell has a recent New Yorker article titled "Most Likely to Succeed: How do we hire when we can't tell whose right for the job." (here). He compares teachers, NFL quarterbacks and financial advisors. He make much the same points.
(2) Dealing with the equity issue (giving poor and minority students good teachers, good schools) and how to get underperforming students to perform better are not enough if what students learn and are tested on are not the most relevant, important skills for their lives. We overemphasize math skills (yes, we need more scientists and engineers, but why algebra for everyone) and we underemphasize world (foreign) language skills. To give world languages the emphasis they deserve we should have many more world language immersion program beginning in kindergarten and our educational testing system should test and recognize student proficiencies.

Gregory said...

I think we need both foreign language proficiency and math--I don't know why this would be an either/or proposition.

The problem with the "great teachers" factor is:
1) it is self-referential--we seek great outcomes by hiring great teachers who, by definition, produce great outcomes. We need better definitions of what are great teachers.
2) it is a lagging indicator, and we may not know who were the great teachers until they are gone. Once again, we need more concise indicators of what makes a great teacher.

Just saying "we need more great teachers" is a meaningless statement.