New model offers districts same-year teacher turnover signals

School systems may no longer have to plan for current-year staffing with teacher-turnover numbers that arrive after the hiring season is effectively over. A new October 2026 working paper from Joshua Bleiberg and Tuan D. Nguyen says it can nowcast district teacher turnover in the current school year by combining two decades of district records with labor-market and economic indicators. Its national estimate suggests turnover rose to 14.6 percent in 2025-26 before easing to 13.2 percent in 2026-27. (edworkingpapers.com)

What changed is not simply another warning that turnover remains high. The paper argues that districts and states can get a same-year signal, plus a peer benchmark, instead of waiting for slower official counts or relying only on survey snapshots. That matters because the latest federal school-level turnover figures from the National Center for Education Statistics still track teachers from 2020-21 into 2021-22, while RAND’s more current national estimates are based on district reports of resignations and retirements rather than administrative turnover files. In other words, leaders today choose among data that are either late, narrower in definition, or both. (learningpolicyinstitute.org)

From lagged counts to current-year signals

Bleiberg and Nguyen say they assembled district-level teacher-turnover data from every state for 2003-04 through 2024-25, then paired those records with education, labor-market, and economic predictors from 2000-01 through 2024-25 to estimate current-year turnover. Instead of offering only a raw percentage, the model also sorts districts into lower, similar, or higher turnover relative to peers with comparable turnover trends. That design choice is important: a district’s 12 percent turnover rate can mean something very different in a labor market where churn has long been elevated than it does in a district that was stable until recently. (edworkingpapers.com)

The paper builds on the same researchers’ 2026 peer-reviewed article in AERA Open on using the U.S. Census Bureau’s Quarterly Workforce Indicators, or QWI, to create valid, more contemporaneous K-12 labor-market measures. In that earlier work, the authors argued that QWI helps address four recurring weaknesses in education labor data: inconsistent definitions, thin local samples, lack of subgroup detail, and the basic fact that many education datasets arrive too late for operational decision-making. Census describes QWI as a set of 32 indicators covering employment flows, wages, hires, and related measures, available across multiple geographic levels including counties and states. (journals.sagepub.com)

That timing problem is not abstract. Research by John Papay and Matthew Kraft found that hiring teachers after the school year starts reduced student achievement, while Lucy Sorensen and Helen Ladd found that high turnover carries lasting costs for instructional quality and student outcomes. If a district can identify elevated churn while contracts, transfer windows, induction staffing, and spring recruitment plans are still in motion, it has a chance to do something other than document the damage later. (jstor.org)

What the model seems best suited to do

The paper’s strongest operational claim is narrower than “we can tell you the exact turnover rate right now.” In validation checks, the authors report that fewer than 1 percent of districts had turnover inconsistent with their lower/similar/higher classification. That suggests the peer-benchmark flag may be the most decision-useful output, especially for central-office teams deciding where to investigate first. A superintendent probably does not need perfect precision to know whether a district is behaving roughly as expected or is breaking away from its own historical pattern and its peers at the same time. (edworkingpapers.com)

That distinction matters for policy. Exact percentages invite false confidence; a classification system invites triage. A district flagged as higher-turnover-than-peers could move faster on stay interviews, onboarding support, principal coaching, vacancy forecasting, or targeted retention incentives. A state agency could use the same signal to decide which districts need technical assistance before the next budget cycle, rather than after students have already spent months with substitutes or late hires. But peer benchmarking also has a trade-off: “similar to peers” does not mean healthy. If turnover is broadly high across comparable districts, the benchmark may normalize churn that is still harmful for students and expensive for schools. (edworkingpapers.com)

Useful for triage, not a diagnosis

Readers should also keep the paper’s limits in view. EdWorkingPapers describes its papers as work-in-progress publications circulated for comment and discussion, not products that have gone through rigorous peer review. And a nowcast is still a model-based estimate, not a final audited administrative count. The especially current 2026-27 figure should therefore be read as a staffing signal, not as settled fact. (edworkingpapers.com)

There is another practical limit: the paper is built at the district level. That makes it potentially powerful for systemwide planning, but it cannot by itself tell leaders whether the real problem sits in special education, secondary math, one high school principal pipeline, or a cluster of hard-to-staff elementary schools. Districts would still need school-, role-, and subject-level follow-up before turning a warning light into an intervention plan. That is one reason the paper may be more useful for budgeting and staffing triage than for precise diagnosis. (edworkingpapers.com)

For context, the most recent federal school-level turnover data show 84.1 percent of public school teachers stayed at the same school in 2021-22, while 7.9 percent moved schools and 7.9 percent left teaching. A March 2026 Learning Policy Institute report likewise noted that its most recent national estimates drew on NCES data from 2020-21 and 2021-22. Those figures are useful benchmarks, but in October 2026 they are already several school years old. The appeal of the new nowcast is not that it replaces official data; it is that it tries to bridge the years between official releases and real staffing decisions. (nces.ed.gov)

The question now is whether districts treat this kind of tool as interesting research or as planning infrastructure. If the later observed counts for 2026-27 end up tracking the nowcast closely, districts may start to expect current-year turnover intelligence the way they already expect weekly attendance or monthly budget updates. If they do not, the paper will still have clarified something important: education systems know they cannot keep managing staffing with rearview-mirror data, and the race is now on to build a version of turnover measurement that is both timely enough to use and trustworthy enough to act on. (edworkingpapers.com)