How much value is added by value added models? An analysis of teachers' performance over time using New York state assessment data

There is a strong movement to evaluate teachers on the basis of students' performance. To compare teachers fairly, as each may have a mixture of students with different abilities in a given subject area, one should account for variables reflective of students' subject knowledge and background when entering a course. Most methods of control consist of highly sophisticated statistical models mostly difficult to explain to educators who are being evaluated using such methods. This research presents two value-added methods that could be replicated by using in-house resources and standardized student assessment data which are either continuous or ordinal. One method is simpler to implement if one's goal is to evaluate teachers' performance based on students' assessments scores reported as ordinal measures. The second method is similar to a more typical value-added approach and uses hierarchical linear structures to determine a classification of teachers' performance based on their students' assessment scores reported as continuous measures. Teachers' "value-added" in a given academic year is typically calculated using students' longitudinal New York State assessment data, reported in both ordinal and continuous forms. Comparison of results obtained from both methods, along with their interpretations, are used to examine trade-offs between accuracy of methods and their ease of use and transparency. The code used is included for practitioners who may wish to replicate this value-added methodology. Suggestions related to educational policy and feasibility of implementation of methods are also discussed.