Academic Growth and School Accountability

Michigan Department of Education · February 25, 2020

Academic Growth and School Accountability — Michigan Department of Education, February 25, 2020.
Published

February 25, 2020

Event
State agency
Slides
7
Format
Converted from PowerPoint

This deck is an image-based copy of the original slides. Builds and animations are flattened; the slide text is reproduced below for search and screen readers.

Slide text
  1. Academic Growth and School Accountability Meeting with the Michigan Department of Education February 25, 2020
  2. Guiding Principles • There is no single ‘gold standard’ for producing measures of academic growth • Decisions are influenced by factors such as: – How will results be used? – What questions do we want to answer? – How will the model support the values and policy priorities? – What are the conditions and constraints that influence implementation? Growth for Accountability
  3. Adapted from Castellano and Ho (2013) There are a lot of choices… Characteristic 1. Gain Score (Raw, 2. Trajectory (Growth 3. Categorical 4. Residual Gain 5. Projection 6. Student Growth 7. Multivariate Simple, Average) to Standard) (Transition Matrix, (Percentile Rank of (Regression Model, Percentile (EVAAS, Layered Value Table) Residuals, Resid Prediction Model) (Conditional Status Model, Cross- Difference, Covariate %ile, %ile growth classified Model) Model) trajectory) Description Describes growth Extends gains or Defines growth by Describes growth as Uses past scores to Percentile rank of Uses entire student with simple average gains in a transitions among the difference predict future scores current status in a score histories, differences or average (usually) linear fashion status categories (e.g., between current status through regression reference group of including other gains over time toward a future Basic, Proficient, and the conditional equations students with similar subjects and teachers, achievement target Advanced) over time mean status given past scores to detect higher than (e.g., proficiency in 3 scores expected student years) scores associated with particular teachers Question Addressed How much has a Does the student’s How has this student How much higher or Given this student’s What is the percentile Is this teacher student learned on an gains exceed the target grown in terms of lower has this student past scores, and based rank of a student associated with higher absolute scale? established based upon transitions through scored than the on patterns of scores compared to students scores for his or her the established categories over time? conditional mean in the past, what is her with similar score students than trajectory to given her past scores? predicted score in the histories? What is the expected given all proficiency. future? minimum SGP a available scores and student must maintain other teacher effects? to reach a target standard? Primary Inference Growth Description Growth Description Growth Description Growth Description Growth Prediction Growth Description Value-Added (gain score) (criterion-referenced (categorical change) (norm and growth to based upon trajectory standard/criterion based standard) referenced) Statistical Basis Gain-Based Gain-Based or can be Gain-Based Regression Regression Regression Regression calculated using other approaches (e.g., regression, SGP) Required Data Vertical scale Vertical scale for gain- Articulated cut scores An interpretable scale. Interpretable future Large sample sizes for For high-stakes value- Features based approaches, no across years and Assumptions of linear scale or future reliable estimation. added, many years of vertical scale for SGP grades. Values for regressions. standard. student data for stable approach value tables. Implicit teacher effects. vertical scale. Group-level Average gain Average trajectory or Average across value Average residual gain Average future Median or mean SGP, Only group-level Interpretation percentage of on-track Growth for Accountability tables or percentages prediction or percentage of on-track interpretations: students of on-track students percentage of on-track students Teacher- and school- students level “effects”
  4. What models are states using for accountability? Growth Model Count Examples Student Growth 23 AL, AZ, CO, DC, GA, HI, IA, IL, IN, MA, Percentiles (SGPs/ MD, MI, NH, NJ, NM, NV, OR, RI, SD, AGPs) UT, VT, WA, WY Value Added 6 AR, MO, OH, PA, TN, SC Model Growth to 4 CT, LA, ME, NE Standard Value Table 4 AK, KY, OK, VI From DQC Growth Data (2019). https://dataqualitycampaign.org/wp- content/uploads/2019/04/DQC-Growth-Data-Resources.pdf Growth for Accountability
  5. Model decisions should be informed by criteria • Policy considerations – Does the selected approach align with the state’s theory of action for the accountability system? – Does the model provide information to inform the highest priority questions? – Are the expectations for ‘good enough’ growth appropriate and reasonable? • Technical considerations – Are results valid, reliable, and fair? – Are relationships with factors thought to be unrelated to growth minimized (e.g. prior performance, school size)? – Are there sufficient and appropriate assessment data to support the model? • Practical considerations – Are the model features open to inspection and customization? – Can it be implemented with current resources? – Are results clear and usable to stakeholders? – Is the model robust to changes in the assessment(s)? Growth for Accountability
  6. Next steps • We suggest investing time to articulate the state’s priorities for growth to develop criteria related to policy, technical, and practical considerations • By so doing, the state can develop a set of criteria to – Guide model selection – Determine model features – Inform ongoing evaluation Growth for Accountability
  7. Some examples of potential criteria: • The state values approaches that: – Provide both norm- and criterion-referenced information – Detect important school level effects – demonstrate availability of the full distribution of growth outcomes to schools of various demographic compositions (e.g., poverty, SWD, and ELL) – appropriate correlations with prior year status – Provide clear and usable results – Are robust to changes in state assessment and differences in test characteristics – Are open and transparent …etc IL Acct TAC Webinar Jan 23, 2018