Describing the Malleability of Teaching Practice: Profiles of Instructional Growth in the Measures of Effective Teaching Project and the District of Columbia Public Schools

• In Event: Promising Scholarship in Education: Dissertation Fellows and Their Research

Abstract

Purpose of study
Research has shown that there are large differences between teachers in their ability to help students achieve, and that these differences have large and persistent effects for students (e.g., Rivkin, Hanushek & Kain, 2005). In recognition of these differences, most districts invest heavily in teacher professional development (PD) which is often targeted toward changing specific areas of instruction. For instance, in the District of Columbia Public Schools (DCPS) where teachers are observed five times a year on the Teaching and Learning Framework (TLF), instructional coaches help teachers work toward attaining proficiency on standards where they have obtained low ratings.

While identifying and targeting specific areas of instruction is an improvement over “one-size-fits-all” traditional models of PD, this method may importantly neglect how practices are interrelated within a teachers’ overall professional practice. In fact, we know very little about patterns of practice, or “instructional profiles.” If we think that there may be groups of teachers who share similar characteristics in their instructional practice (e.g., weak in ability to provide strong content explanations, but strong classroom management), it may be helpful to be able to explicitly identify these profiles in order to better coordinate professional development across multiple teaching competencies.

Moreover, it may also be useful to identify profiles of instructional growth. Though we know teachers improve their contribution to student test score outcomes (i.e., value-added; Atteberry, Loeb & Wyckoff, 2013; Papay & Kraft, 2015) through the course of their early career, we don’t have the same empirical evidence on how teachers’ instructional practice develops over time or in the presence of formative feedback from an ongoing evaluation system. Many scholars have contributed theoretical frames for teacher development (Berliner, 1988; Borko & Livingston, 1992; Grossman, 1990; Grossman, Smagorinsky & Valencia, 1999; Shulman, 1987), but until recently the dearth of longitudinal data on classroom performance has made it impossible to explore theorized patterns of instruction or development empirically on a large scale.

The goal of this project is to explore profiles of instruction and instructional growth using data from classroom observation in two contexts: DCPS’s IMPACT teacher evaluation system and the Measures of Effective Teaching (MET) project. Specifically, I will address the following research questions:

Research Questions
1. What instructional profiles do we observe in DCPS using the TLF and in the MET study using the Framework for Teaching?
2. What instructional growth profiles do we observe?
3. How predictive of profile classification are school and individual characteristics (e.g., school poverty level and teacher experience)?

Analytic Approach
This analysis uses person-centered latent variable approaches to identify instructional profiles and instructional growth profiles using classroom observation data. First, I employ a finite mixture model, specifically Latent Class Analysis (LCA), in cross-sectional data to investigate how teachers cluster in the data in terms of their classroom observation indicator scores (Collins & Lanza, 2010; Halpin & Kieffer, 2015; Muthén & Muthén, 2000). Similar to factor analysis, LCA attempts to understand latent (unobserved) constructs using the covariance structure of measured (observed) item-level data. Unlike factor analysis, which looks for variables that group together in the data, LCA looks for individuals who are clustered together, forming homogenous subpopulations within a larger heterogeneous population that may not be adequately described by population means. Extending this technique, subpopulations can also be identified based on similar growth trajectories (Muthén & Asparaouhov 2000; Jung & Wickrama, 2007). This class of techniques, known as Growth Mixture Models (GMMs), improves upon conventional growth models by allowing the estimation of unique growth and variance parameters for empirically-derived classes of growth.

Policy Implications and Preliminary Result
The goal of this study is to provide an initial, individual-centered descriptions of instruction and instructional growth using detailed longitudinal classroom observation data. These profiles could have important implications for current approaches to teacher professional development, coaching, and evaluation. If meaningful typologies of teaching or teacher growth emerge, these could be used diagnostically to to provide more targeted learning opportunities based on descriptions of these profiles.

Author

• Melinda K Adnot, University of Virginia