Private school effects in urban and rural India: Panel estimates at primary and secondary school ages

Abstract

I present the first value-added models of learning production in private and government schools in India using unique panel data from Andhra Pradesh state. I examine the heterogeneity in private school value-added across different subjects, urban and rural areas, medium of instruction, and across age groups. Further, I also estimate private school effects on children's self-efficacy and agency. In rural areas, I find a substantial positive effect (>0.5 SD) of private schools on English, no effect on mathematics and heterogeneous effects on Telugu for 8-10 year old students; at 15 years, there are modest effects (<0.2 SD) on mathematics and Telugu receptive vocabulary. I find no evidence of a positive effect in urban areas or on psychosocial skills. Results on comparable test domains and age groups correspond closely with, and further extend, estimates from a parallel experimental evaluation.

Estimating Private School Effects for School Children in Peru: Evidence from Individual‐level Panel Data

“…The main idea behind the aforementioned specification is that lagged achievement will capture the contribution of all previous inputs and any past unobservable endowments or shocks. This is the structural assumption of the value‐added models, and it is generally believed to be a significant improvement over a contemporaneous specification, which explains current achievement using only current inputs (Andrabi et al, ; Singh, ; Todd & Wolpin, ). Equation rests on the assumptions of geometrical decay of both mental endowment and inputs, and the sufficiency of lagged achievement as an identification statistics.…”

“…In this paper, we adopt the dynamic OLS or lagged value‐added model for our analysis. In writing down the empirical specification used in estimating private school effects for 12‐year old children in Peru, we follow the sequential approach by Singh () and write the following: Private it represents an indicator variable that takes on the value of 1 if a child is enrolled in private school in 2013 and 0 if enrolled in public school; Y it − 1 is lagged achievement (test scores) measured at baseline; X represents a vector of contemporaneous background characteristics that include cluster dummies, paternal and maternal level of education, paternal and maternal age, wealth index, household size, distance to school, child's height‐for‐age, child's body mass index and the sex of the child; time it captures the number of hours spent on various activities in a typical day. The time use controlled for include time spent in school, time spent studying outside school, time spent caring for family/relatives, time spent on domestic task, time spent on farming activities, time spent on household chores and time spent on paid job.…”

Cost-Effectiveness Analysis in Development: Accounting for Local Costs and Noisy Impacts

“…Standardized tests usually report results in standard deviations, but the size of a standard deviation is dependent on the variation in the underlying population, and there is no reason to believe that will be consistent across populations. Furthermore, different methods of weighing individual question scores to reach an aggregate score in standard deviations still make for a great lack of comparability (Singh, 2015).…”

Disrupting Education? Experimental Evidence on Technology-Aided Instruction in India

“…While standard in the international assessments literature for generating comparative test scores, the use of IRT models is much less prevalent in the economics of education literature in developing countries (for notable exceptions, see Das and Zajonc 2010, Andrabi et al 2011, Singh 2015. For a detailed introduction to IRT models, please see Van der Linden and Hambleton (1997) and Das and Zajonc (2010).…”