National and Labour are offering quite different views on what new policies will improve the performance of the school system.  Labour believes that the current standard of teachers is fine, but what is needed is more of them.  National believes that improving school outcomes will come from improving the average standard of teaching.  Their proposal is to reward better teachers and, in particular, reward them for assisting other teachers to improve education outcomes. 

Both approaches have received their share of criticism.  National’s approach has been characterised as skimming money away from students and into the pockets of senior teachers and Labour’s as a mechanistic response that ignores the quality of teachers.  So who is right?  Will we get more bang for the education buck for having more teachers or from having better teachers?

The international evidence increasingly suggests that it is National that is on the right track and that the best way to improve education outcomes is to improve teacher quality.  To begin with, the OECD sponsored PISA study of international education attainment highlighted that the largest variation in student outcomes was between classrooms, as distinct from between schools or between students within the same class.  This evidence has generally been interpreted as meaning that the quality of teachers really matters.

A couple of papers recently published in the American Economic Review by Raj Chetty, John N Friedman and Jonah E Rockoff reinforce the view that teacher quality is important for education outcomes.  Their first paper addresses the information content in estimates of value added measures of student achievement.  That is, do differences in student improvements in test scores between teachers capture differences in teacher performance?  

The authors use a dataset on test scores and teacher assignments in “grades 3-8 from a large urban school district in the United States”.  The data covers more than 2.5 million students and includes over 18 million tests of Maths and English during the period from 1989 to 2009.  Further they match 90% of the observations in the school district data with US tax records from 1996 to 2011.  With this rich dataset they are able to separate many of the student-specific influences on test results (eg natural aptitude and their family income base) from the influence of teachers.  

By accounting for family characteristics they find very little evidence of sorting (ie self-selection) of students influencing teacher value added measures.  There appear to be two reasons for this result, the first is that parental characteristics impact more on the average performance of students than on the annual increment in student performance, which is the basis of teacher value added measures.  A second reason for why sorting is so limited seems to be due to 85% of the variation in teacher value added being within, rather than between, schools: since most sorting occurs through the choice of school, parents have less scope to steer their children toward higher value added teachers.

The authors also adopt a second approach, which was to follow teacher arrivals and departures from schools to test whether changes in student scores are consistent with the change in teacher profile.  That is, if an apparently high value add teacher leaves a school one would expect that the fall in average teacher quality would lower subsequent test scores.  They found that forecast changes in test scores based on teacher changes closely matched observed changes.  Based on these results they concluded that test scores, particularly when compared with previous year results, provide unbiased estimates of teacher performance.

Having established that student test scores are indeed a good measure of the value added provided by teachers, the second paper addresses the deeper question: does having high value add teachers improve student long-term outcomes?  Using the same database described above, the authors were able to track about one million individuals from primary school to early adulthood, where they measured outcomes such as earnings, college attendance, and teenage birth rates.

They find that teacher ability has substantial impacts on a broad range of outcomes.  Having better teachers during primary school:

They also found that teacher ability has slightly larger long-term impacts for female students than for male students.  Also that improvements in English teacher quality have larger impacts than improvements in Maths teacher quality.

Finally their results indicate that there are likely to be greater gains from raising teacher standards at the bottom end of the quality scale.  They estimate that replacing a teacher who is in the bottom 5% with an average teacher would increase the mean present value of students’ lifetime income by $US250,000 per classroom over a teacher’s career, a gain that they estimate is likely to be more than ten times the expected additional cost of compensating teachers for the risk of evaluation based on value added measures.

Conversely, they are less positive about the benefits of paying bonuses to high quality teachers in order to increase retention rates, mainly because high quality teachers already have high retention rates.  But they do note that high salaries could attract more high quality teachers to the teaching profession, but they do not investigate this issue.

These last points suggest that there is an element of glass half full and half empty from National.  Better remuneration for better teachers is unlikely to have a big impact on teacher retention (but may have some impact on attracting better teachers to the profession).  However, encouraging good teachers to help raise standards where challenges are greatest would appear to be a high value use of the nation’s teacher resource and in line with the evidence collected by Chetty, Friedman and Rockoff. 

This evidence provides more support for the recommendations by Treasury in their 2011 briefing to the incoming Minister of Finance (p21):

New Zealand’s compulsory education system produces good outcomes for most students, as evidenced by our strong performance in international tests.  However, despite large funding increases, achievement levels remain unacceptably low for some groups.  Student achievement can be raised by improving the quality of teaching, which the evidence shows is the largest in-school influence on student outcomes.  Increasing student/teacher ratios, and consolidation of the school network, can free up funding that could be used to support initiatives to enhance the quality of teaching, such as more systemic use of value-add data and a more professionalised workforce.

This recommendation was obviously influential in influencing National’s approach to education policy. 

The approach of Labour appears to be at odds with Treasury’s perspective and indeed seems to miss some of the subtleties in the evidence that they cite in support of their policy to reduce student teacher ratios.  The Labour education policy quotes papers by both David Zyngier and Diane Schanzenbach in support of their stress on class size. 

The first point to note is that both of these papers were written to counter arguments and evidence that suggest that class size is not a critical issue in education outcomes.  So although class size might yet help education outcomes, there is not the “consensus amongst education researchers” that Labour claims. 

In addition both Zyngier and Schanzenbach highlight that teacher standards are important for the attainment of results from reducing class sizes.  The potential benefits of smaller class sizes diminish if teacher standards are not maintained.  Further, it is intriguing that Zyngier stresses that smaller classes are likely to have their most profound impact during the first four years of school.  These are the years when children benefit most from one-on-one tuition time, yet Labour’s class size policy is aimed at year 4 and above, ie after the years when small class sizes are likely to have their biggest impact.  

One is left wondering how deep was the thinking behind the Labour policy and how focussed they were in developing a policy that would actually deliver better education outcomes for students.  Are there perhaps other reasons for wanting to increase teacher numbers?