Do teachers get better with experience?

Do teachers get better with experience? This is a question that we are seeking to answer as part of our teacher improvement through data and evaluation project (TIDE), funded by the Nuffield Foundation. The answer to this question has implications for deployment, pay scales, class allocation, recruitment and retention incentives, and professional development. But what can the literature tell us? And how does Matthew Wiswall’s work on teacher effectiveness-growth impact our conclusions?

Where to start?

We began answering this question by looking at the evidence in the literature; given how important this is, one might imagine that the science would already be airtight, and that we would have good evidence for how teachers improve with experience, for different subjects, phases, and for different periods of their careers.

There have been several reviews of the literature, and these have concluded that: (one) teachers' effectiveness grows in the first one to five years of their career and (two) that this effectiveness growth slows down or stops entirely thereafter. However, the evidence suggests that while a new teacher does get better with experience, on average, they do not typically improve so much as to close the gap with the top performers, solely through experience. For example, if an average first-year teacher is, hypothetically, at the 40th percentile of effectiveness initially, after five years they might move to the 50th or 60th percentile with that ~0.1 SD gain.

But work by Matthew Wiswall in 2013 upended this by showing that the evidence that teachers' effectiveness plateaus after five years is an artifact of the analysis approaches that other researchers had been using. It is worth taking some time to examine Wiswall's work and think about the implications for policy questions in England.

Unpacking Wiswall’s work

Wiswall used data from elementary schools in North Carolina, focusing on grade five teachers (teachers tend to specialise in a year group in the US), spanning 10 years from 1996 to 2005. He used maths and reading test scores to estimate the value-add for each teacher and then how this changes over time in teaching.

Using the models that others had used in the past, Wiswall was able to replicate the result that the teachers were improving a lot in the first one to five years and then not much at all after this. But he noticed a quirk of the models that others had used: they grouped the years after five years together such that it was nearly impossible for the data to show a change over time. When he made the model more flexible, comparing teachers' effectiveness with the baseline throughout their careers, he found that there was a relatively simple upwards slope for maths outcomes, but still more of a plateau for reading outcomes. This implies that the conclusions from other studies might have been baked in from the outset simply because of the way experience was modelled, and that perhaps teachers can continue to improve, at least in maths.

Although he found that teachers can improve with time, Wiswall also showed that more experienced teachers are not, on average, more effective than inexperienced teachers. But given what was shown about teachers improving with time, how can this be true?

The leaving-teacher effect

The second big contribution of Wiswall’s paper, showed that the more effective a teacher was when they started teaching then the more likely they were to stop teaching in the North Carolina public school system in any one year after that. Astonishingly, this tendency for the more effective teachers to leave exactly balanced out the extent to which the teachers who remained improved in their teaching. For example, if in a year, 10% of teachers leave; then, according to Wiswall, they would be better teachers on average than the 90% who stayed, but those 90% who stayed would have improved just enough in the year to offset the loss in above-average effectiveness in the 10% who left.

See the figure below, taken from the paper: the dark line shows that average teacher effectiveness is flat; the ‘average teacher FE [fixed-effect]’ line shows the declining average teacher effectiveness in the cohort as measured when they started teaching, and the ‘return to experience’ line shows how much teachers improve so long as they stay in teaching.

To summarise

According to this paper, elementary teachers do continue to get better at teaching maths, but since teachers who were more effective when they started are more likely to leave, the result is that at any one time teachers who have more experience are only ever so slightly more effective than those that do not. Which raises the question: how much better would maths teaching be in the school system that Wiswall was looking at if everyone stayed in teaching until they retired? Quite a bit better: based on this model, grade five maths outcomes would be one SD higher on average if teachers stayed on in the profession, all else being equal. This is a big effect, equivalent to a year's progress according to the Education Endowment Foundation’s (EEF) translation from effect sizes to months progress.

Conclusions and takeaway

We can be pretty confident that on average teachers become more effective over the first one to five years, but unsure about what happens after that; Wiswall’s paper from 2013 has suggested that teachers do continue to improve, at least in maths, and that other research has methodological flaws which have led them to miss this. If so, then keeping teachers in teaching could have substantial impacts on pupils’ outcomes. But more generally, the evidence on such an important topic is thin.

However, this evidence of teacher effectiveness-growth is only loosely applicable to English schools, especially secondary schools. For a start, almost all of the evidence available comes from the US. This is clearly problematic for making inferences about England, as teachers in the US are trained and deployed differently. But worse still, the evidence from the US uses data from 'elementary schools' (primary schools) only, and there is almost no evidence regarding teacher effectiveness growth from secondary schools (in the US or otherwise). Conclusions for secondary English schools rely on a double leap: from US to England, and from primary to secondary. Since England and the US are different, and teaching in primary and secondary is different, each could be a leap too far. Furthermore, while the impact of teachers on academic attainment is important, it is a narrow measure of teachers’ effect, and changes in effect on attainment is a shallow description of the ways that teachers will develop in their roles.

More research is needed in England to replicate Wiswall's work, this time with English primary and secondary-school teachers, to inform decisions with the appropriate evidence. With funding from the Nuffield Foundation, we are answering these questions as part of the TIDE research project. We are using schools’ data to model returns to experience, having shown that teacher value add can be estimated using the data. The next step is to build on the work of Wiswall and others to generate insights into teacher effectiveness in English schools.

Learn more about the TIDE project and teacher education dataset (TED). Keep track of project updates via the TIDE and TED blog and news page.