Learning equity outcomes and policies in India: Simulation Ginis and learning outcomes results

• In Event: Improving learning equity outcomes and policies: Concepts, methods and empirical findings from India, Nigeria and Malawi.

Proposal

UN SDG4 has helped to focus international and national attention on improving the quality of education—and on learning. This has led to substantial increases in attention to, and international development assistance towards, the improvement of educational quality worldwide.

Learning equity research (each based on a particular population group) can enhance our understanding of the disparities in learning progress over time among different marginalized segments of a country’s population, such as by socio-economic status, age, gender, language, ethnicity, disability, geography and so forth. Such work allows for comparisons across varying learning outcomes by assuring that learners within a specific population are measured on a consistent scale. It will also provide reliable ways to evaluate the impact of targeted interventions to improve learning over time.

In the present research, we sought to better understand how the distribution of EGRA data in low-income countries could serve to simulate interventions that would impact various parts of the reading outcomes distribution. Through this simulation one can observe how varying distributions of learning scores relate to a change in Lorenz curves, a measure of learning equity.

In the first simulation (Simulation A) there were three learning scores: LS1, LS2, and LS3. Learning Score 1 (LS1) “mimics” the learning scores in oral reading fluency from a particular chosen country (Dubeck & Gove, 2018), with a mean of 20 points (or 20 correct words per minute). LS2 and LS3 represents, consecutively, average 30 or 40 points. These three distributions allowed for the calculation of Gini Learning Index (GLI) scores of each dataset in order to observe how an increase in students’ score at the bottom of the pyramid is associated with greater equity in learning scores.

Simulation B demonstrates how varied inputs would impact learning outcomes. Hence, Simulation B uses Learning Score 1 (LS1) taken from Simulation A to predict the additional impact of input variables – such as education resources (some malleable as in improved teacher training, textbooks, language ‘match’ between child and teacher, etc.) and positive demographics (some non-malleable as in caste or religious status, parental education, resources at home) – to predict improved learning outcomes.

Currently, research and analysis is ongoing in India, in partnership with a major Indian NGO and the Indian Ministry of Education, using datasets (N= approx 5,000) in Uttar Pradesh over several several years to determine the validity of Simulations A and B to make predictions about reducing the GLI (increasing learning equity) within varied contexts and population samples of primary school children.

Learning equity research can provide a particular window on within-country learning improvements, with information that can help local stakeholders and teachers better target their instruction. The paper concludes with thoughts on how to manage this long-standing challenge of implementation learning policies that matter locally when there are pressures to reach national or international standards.

Authors

• Daniel A Wagner, University of Pennsylvania

• Amrit Thapa, University of Pennsylvania