How Did Teachers’ Effectiveness Change When Instruction Moved From In-Person to Remote Learning?

Lawson and Sass (2026) study how teachers’ effectiveness changed when schools shifted from in-person to remote instruction. They ask whether the move to online learning altered relative teacher performance and which teacher traits predicted success. They analyze matched student–teacher administrative data from three large metro-Atlanta districts, using fall-to-winter math and reading test score growth for grades K–8. They find that variation in teacher effectiveness increased during remote learning, and the year-to-year correlation in value-added fell to 0.11 compared with roughly 0.30 pre-pandemic. Veteran teachers were modestly more effective. About 37% of teachers in the top quintile before remote learning dropped to the bottom two quintiles. Early-career teachers were overrepresented in the lowest decile (14.1%).

Why This Article Was Selected for The Policy Scientist

This research article address a policy question of enduring significance: how instructional modality reshapes the distribution of teacher effectiveness. As remote and hybrid learning become structural features of K–12 education rather than emergency measures, understanding performance variation across modalities has implications for staffing, professional development, and equity. The study is timely given persistent virtual offerings and ongoing debates about learning loss and instructional quality. It contributes to the literature by extending value-added research into remote contexts, building cautiously on prior work linking teacher characteristics and effectiveness. The administrative student–teacher dataset is rich and longitudinal, though geographically concentrated. Findings are plausibly informative for similar districts. The statistical approach is rigorous; future work using stronger causal designs would further sharpen inference.

Full Citation and Link to Article

Lawson, M. C., & Sass, T. R. (2026). Teacher effectiveness in remote instruction. Education Finance and Policy. https://doi.org/10.1162/EDFP.a.437

Central Research Question

This study examines how teachers’ relative effectiveness changed when instruction shifted from traditional in-person classrooms to remote learning during the COVID-19 pandemic. The authors ask two closely related questions. First, did the transition to remote instruction alter the distribution of teacher effectiveness, including the stability of teacher performance rankings across years? Second, which observable teacher characteristics, particularly years of experience, were associated with stronger or weaker performance in a virtual environment? Rather than focusing on student learning loss alone, the analysis centers on whether the instructional modality itself reshaped patterns of teacher value-added. The research therefore addresses a fundamental policy concern: whether teacher effectiveness is context-dependent and whether the skills associated with successful in-person teaching translate to remote settings.

Previous Literature

Prior research has extensively documented variation in teacher effectiveness in conventional classroom environments. A large empirical literature using value-added models shows that teachers differ meaningfully in their contributions to student achievement growth. Studies have also established several regularities, including modest year-to-year stability in teacher value-added estimates and systematic improvements early in teachers’ careers. Teacher experience, while not uniformly predictive, is frequently associated with gains in effectiveness, especially during the first several years of teaching.

Evidence on remote instruction, however, has been comparatively limited. Earlier studies of online education often relied on surveys, self-reports, or postsecondary samples, emphasizing pedagogical practices rather than measured effectiveness. Research on virtual charter schools identified negative student outcomes and differences in teacher characteristics but rarely tracked the same teachers across instructional modalities. The present study contributes by linking the established value-added framework with the unique natural experiment created by widespread pandemic-era remote learning. It extends the literature on contextual variation in value-added estimates, complementing work that has examined differences across school types, student populations, testing instruments, and institutional settings.

Data

The analysis uses matched student–teacher administrative data drawn from three large school districts in the metro-Atlanta region. The dataset includes student-level test scores from nationally normed formative assessments in math and reading, administered multiple times within each academic year. Scores are standardized by grade and year relative to pre-pandemic national norms, permitting comparisons across cohorts. Student demographic characteristics are available, including gender, race/ethnicity, disability status, English learner designation, and indicators of economic disadvantage where reported.

Teacher-level information includes gender, race/ethnicity, and years of experience. The panel structure allows the authors to follow both students and teachers over time. Crucially, the districts differed in their instructional policies during 2020–21, with two districts operating under fully remote instruction in Fall 2020 and one district implementing a hybrid model with varying shares of remote teaching. This variation enables comparisons between semesters of universal in-person instruction (Fall 2019) and semesters characterized by remote delivery (Fall 2020). The data also allow for robustness checks addressing testing location, sample attrition, and consistency of teacher observations.

Methods

Teacher effectiveness is estimated using value-added models that measure a teacher’s contribution to student test score growth within a semester. Specifically, winter test scores are modeled as a function of prior fall scores, student demographic characteristics, class size, and teacher fixed effects. Estimation is conducted separately by district, subject, and year, ensuring that value-added estimates are interpreted relative to the average teacher within each context. This cross-sectional strategy allows the relationship between covariates and achievement growth to vary across instructional modalities.

To address noise inherent in value-added estimation, the authors apply multiple restrictions. Only students taught by a single instructor of record in a subject are included. Classes with very small numbers of test scores are excluded, and teachers with insufficient observations do not receive value-added estimates. Empirical Bayes shrinkage is employed to reduce estimation variance by pulling imprecise estimates toward the mean. The authors also conduct extensive robustness checks, including analyses without shrinkage, controls for test modality, exclusion of rookie teachers, and adjustments for unobserved student heterogeneity using auxiliary growth measures.

Findings/Size Effects

The study reports several notable findings. First, the year-to-year correlation of teacher value-added declines sharply when transitioning from in-person to remote instruction. Pre-pandemic correlations between adjacent in-person years range from approximately 0.28 to 0.35. In contrast, the correlation between Fall 2019 (in-person) and Fall 2020 (remote) falls to 0.11. This magnitude indicates substantially reduced stability in relative teacher performance rankings under remote instruction.

Second, the variance of teacher value-added increases during remote learning. Distributional analyses reveal a wider spread, particularly at the lower tail. Some teachers exhibit effectiveness estimates below the minimum observed in prior in-person semesters. Statistical tests reject equality of variances, confirming that dispersion increased beyond what would be expected from sampling variability alone.

Third, teacher experience is associated with differences in remote performance. Veteran teachers appear modestly more effective, while early-career teachers are disproportionately represented among the least effective instructors. Approximately 14.1% of early-career teachers fall into the lowest decile of effectiveness, compared with 9.4% of mid-career and 8.6% of late-career teachers. These proportions exceed the 10% benchmark expected under a uniform distribution.

Fourth, effectiveness transitions across quintiles show increased downward mobility among previously high-performing teachers. Nearly 37% of teachers ranked in the top quintile of effectiveness in Fall 2019 move to the bottom two quintiles in Fall 2020. This contrasts with roughly 23% during the baseline comparison between two pre-pandemic in-person years. The magnitude suggests that some teachers who were highly effective in conventional classrooms experienced marked declines in relative performance under remote conditions.

Fifth, heterogeneity across grade levels is evident. The increase in variance is concentrated among early elementary grades (K–2), where remote instruction coincides with especially large dispersion in effectiveness. By upper elementary grades, differences between instructional modalities narrow. Middle school teachers show comparatively smaller increases in variance and, in some cases, relatively stronger performance.

Finally, analyses exploiting within-year transitions (Fall 2020 remote versus Spring 2021 return to in-person) indicate that variance contracts once students return to classrooms, despite the pandemic remaining ongoing. This pattern provides suggestive evidence that the widening of effectiveness distributions is linked more closely to instructional modality than to generalized pandemic disruptions.

Conclusion

The findings demonstrate that teacher effectiveness is sensitive to instructional context. Remote instruction is associated with reduced stability in teacher rankings and increased dispersion in effectiveness estimates. The results imply that skills underlying successful teaching may not transfer uniformly across modalities. Observable characteristics such as teacher experience exhibit modest predictive power, with less-experienced teachers performing relatively worse during remote instruction.

The study advances the teacher effectiveness literature by explicitly examining cross-modal variation using longitudinal administrative data. The methodological framework is rigorous and transparent, though causal interpretation remains limited by the observational design. While the pandemic created exogenous shifts in modality, teacher assignment to specific remote environments is not randomized. Future research employing stronger causal inference strategies could refine estimates of modality-specific teacher effects.

Overall, the analysis provides empirical evidence that instructional delivery formats can reshape patterns of measured teacher effectiveness. These results are relevant for policy discussions concerning virtual schooling, hybrid models, teacher evaluation, and workforce development in evolving educational environments.