In Simple Words
Evaluating a runner's stamina using a multiple-choice quiz about running technique instead of measuring actual running distance introduces measurement noise that causes annual test scores to fluctuate unpredictably.
Observed middle school teacher records
- Math scores
- Year-to-year stability: 0.512
- English scores
- Year-to-year stability: 0.117
2008–10 teacher records. Higher regression values mean more stable ratings; the test-format explanation was speculative.
See source →Research question
State English Language Arts (ELA) value-added metrics display substantially higher year-to-year volatility and lower predictive stability than mathematics metrics, reducing their standalone reliability in teacher evaluation systems.
Participants
782 elementary (grades 4–5) and 559 middle school (grades 6–8) math and ELA teachers across 6 urban public school districts.
Study Design
Cross-year linear regression comparing 2009–10 value-added scores against 2008–09 value-added scores across subjects.
Results
ELA value-added stability coefficients between consecutive years were only 0.350 in elementary and 0.117 in middle school, compared to 0.396 and 0.512 in math. This instability stemmed from state ELA tests relying primarily on multiple-choice reading comprehension items that fail to directly capture teacher impacts on student writing.
Takeaway
Apply larger empirical Bayes shrinkage factors and integrate direct open-ended writing assessments when evaluating reading and language arts instructors.