You’ll often hear the argument that half or almost half of all beginning U.S. public school teachers leave the profession within five years.

The implications of this statistic are, of course, that we are losing a huge proportion of our new teachers, creating a “revolving door” of sorts, with teachers constantly leaving the profession and having to be replaced. This is costly, both financially (it is expensive to recruit and train new teachers) and in terms of productivity (we are losing teachers before they reach their peak effectiveness). And this doesn’t even include teachers who stay in the profession but switch schools and/or districts (i.e., teacher mobility).*

Needless to say, some attrition is inevitable, and not all of it is necessarily harmful, Many new teachers, like all workers, leave (or are dismissed) because they are just aren’t good at it – and, indeed, there is test-based evidence that novice leavers are, on average, less effective. But there are many other excellent teachers who exit due to working conditions or other negative factors that might be improved (for reviews of the literature on attrition/retention, see here and here).

So, the “almost half of new teachers leave within five years” statistic might serve as a useful diagnosis of the extent of the problem. As is so often the case, however, it’s rarely accompanied by a citation. Let’s quickly see where it comes from, how it might be interpreted, and, finally, take a look at some other relevant evidence.

The primary source for the claim seems to be analyses by respected University of Pennsylvania professor Richard Ingersoll (presented, among other places, in this 2003 report). Ingersoll uses data from the 2001-02 Teacher Follow-Up Survey (TFS). The TFS is a supplement to the Schools and Staffing Survey (SASS), a highly regarded national survey of teachers conducted by the National Center for Education Statistics (NCES).

And, indeed, he estimates cumulative five-year attrition at 46 percent, and also reports finding similar results using SASS/TFS data from previous years.

So, there is an empirical foundation for this claim, but there are also a few very important caveats and limitations (all of which Ingersoll dutifully acknowledges and/or addresses directly with additional analyses). First, this figure is only an approximation. Few surveys follow respondents for five years, and even fewer contain a sufficient number of new teachers to get precise estimates. As a result, “true” estimates of new teacher attrition are very hard to come by.

The SASS samples a large group of teachers, but contacts a sub-sample of them only one year later (that’s the TFS). This means Ingersoll must derive his five-year attrition figure from estimates of the probability of teachers leaving in their first, second, third, fourth and fifth years.

So, given that the 46 percent is only an approximation, it’s best to give it a little bit of an error margin, and say that, according to this analysis, between 40-50 percent of new teachers left within five years.

Second, due to data limitations, these estimates cannot account for teachers who leave teaching and then return at a later point. This phenomenon is quite common – in fact, some studies estimate that as much as 20-25 percent of leavers return to the profession at some later point (also see here). Some also take non-teaching jobs within the education field.

The fact that many leavers come back does not necessarily detract much from the policy importance of the 40-50 percent figure – schools that lose a teacher must still incur the multi-dimensional costs of hiring a new one, regardless of whether the first teacher eventually returns. (The same basic point of course applies to teacher mobility – e.g., teachers who switch schools/districts.)

On the other hand, a large proportion of returners may mitigate the harmful aggregate effects of new teacher attrition, since many “new” hires will have some experience under their belt, and fewer brand new teachers would be needed.

The third issue is very important, and it usually applies to estimates for a large group (in this case, for a whole nation): There is a tremendous amount of underlying variation. New teacher attrition varies within and between states. In some districts, the rate of leaving (and mobility) is much lower than the national figures suggest, whereas in others, particularly in lower-income areas, it is higher.

For example, one study using Philadelphia data found that, among the teachers hired for the 1999-2000 school year, a full 70 percent had left the district within six years (though it’s very important to note that this analysis, like most single-district estimates of new teacher attrition using administrative [usually payroll] data, significantly inflates the amount of leaving per se, since it doesn’t differentiate between teachers who actually exited the profession entirely versus those who simply got a job in another district). High rates (with the same caveat about inter-district mobility) can also been found in other urban districts, such as New York City and Chicago (also see Ingersoll’s work in this area).

In other words, any national estimate of new teacher attrition (and mobility) might over- or understate the prevalence, depending on the context. Attrition rates are also higher among teachers entering the field with less preparation and mentoring (but there’s some evidence that the former may be more likely to change schools)

The fourth issue to keep in mind about this particular finding is that, by itself, it lacks a frame of reference. In other words, one might point out that the fact that a given proportion of new teachers leave within five years doesn’t by itself tell us whether it’s “high” or “low.” It might be compared with something, such as the rates in other professions.

Certainly, new teacher leaving is higher than that among entrants into other professions, such as law and medicine, that require extensive investment in occupation-specific human capital. But research comparing teachers with a wide range of other jobs is somewhat limited. This 2001 paper looks at 1992-93 college graduates who were teaching in 1994. The analysis indicates that the proportion who had left teaching by 1997 was similar to or lower than those of other graduates’ occupations.**

Ingersoll compared total turnover (attrition plus mobility) with data from other sources, and found that it was higher than that in other occupations by small margins (also see here for a comparison of total teacher turnover [attrition plus mobility] with that of nurses, social workers and accountants). Other studies find an important role of childbearing in contributing to young teachers’ leaving the workforce.

Fifth and finally, these are older data, and even though Ingersoll has stated that the results using the 2004-05 TFS were roughly the same, there is some even newer evidence on beginning teacher attrition that bears on this discussion. NCES, which conducts the SASS/TFS, has begun a supplemental survey survey that follows a cohort of new teachers over time. The first round of results, though preliminary, indicates that, among beginning teachers hired in 2007-08, about 10 percent had left teaching by 2008-09, and only 12.5 percent had left by 2009-10. These are considerably lower than all previous figures, especially the second-year rate.**

On the other hand, the most recent estimates from the TFS (2008-09) find that the raw leaving rate among teachers with 1-3 years of experience is about nine percent, while the rate for teachers with 4-9 years is around eight percent. These figures, very roughly ballparked, are more in line with previous work (though, again, many leavers do return, and it’s the 2009-10 estimate that really diverges).

In any case, it’s obviously true that that job attrition rates can change over time, and it’s quite possible that new teacher attrition may not be as high as it used to be. For instance, comparing the rates from the 2008-09 TFS with those from the previous administration (2004-05), there seems to have been a decrease in leaving among teachers with no prior experience. The recession might have played a role here, as fewer teachers are being hired, and those that do enter may be less likely to incur the risk of changing jobs when so few are available. Shifts in new teachers’ demographics or qualifications could also affect leaving rates.

So, overall, it’s fair to say that the “almost half of new teachers leave within five years” statistic has some backing, but, like almost any standalone statistic, it’s an incomplete assessment of the current situation. To the degree that the approximations from the 1990s and early 2000s are on target, about 40-50 percent of new teachers left the profession within five years, but many did return, and, perhaps most importantly, attrition was higher in some places and lower than others. In addition, there is very tentative recent evidence that the five-year cumulative rate may be somewhat lower going forward.

Regardless, none of this should distract from the larger, important point: New teacher attrition (and total turnover) are serious problems that might be productively addressed. The entire nation may not lose quite as many as half its new teachers, but many districts, especially the poorer, lower-performing ones, do lose half or more, sometimes more quickly than five years. Moreover, if you include mobility (which, from a district’s perspective, has the same basic effect), these rates get much worse.

High turnover among new teachers can create a self-reinforcing cycle that threatens the stability and efficacy of schools, especially those serving the most disadvantaged children. And that’s much more important than quibbling over the precise national rate and how it was calculated.