Here Today, Gone Tomorrow? Examining the Extent and Implications of Low Persistence in Child Learning Tahir Andrabi Jishnu Das Asim Ijaz Khwaja Tristan Zajonc ∗ First Draft: April 18, 2007 † This Draft: January 2, 2009 Abstract Learning persistence plays a central role in models of skill formation, estimates of edu- cation production functions, and evaluations of educational programs. In non-experimental settings, estimated impacts of educational inputs can be highly sensitive to correctly spec- ifying persistence when inputs are correlated with baseline achievement. While less of a concern in experimental settings, persistence still links short-run treatment effects to long-run impacts. We study learning persistence using dynamic panel methods that ac- count for two key empirical challenges: unobserved student-level heterogeneity in learning and measurement error in test scores. Our estimates, based on detailed primary panel data from Pakistan, suggest that only a fifth to a half of achievement persists between grades. Using private schools as an example, we show that incorrectly assuming high per- sistence significantly understates and occasionally yields the wrong sign for private schools’ impact on achievement. Towards an economic interpretation of low persistence, we use question-level exam responses as well as household expenditure and time-use data to ex- plore whether psychometric testing issues, behavioral responses, or forgetting contribute to low persistence—causes that have different welfare implications. JEL Classifications: I21, J24, C23, O12, H4, Keywords: education, fade out, learning persistence, value-added models, dynamic panel data, private schools. ∗ tandrabi@pomona.edu, Pomona College. jdas1@worldbank.org, World Bank, Washington DC and Center for Policy Research, New Delhi; akhwaja@ksg.harvard.edu, Kennedy School of Government, Harvard University, BREAD, NBER; tristan_zajonc@ksgphd.harvard.edu, Kennedy School of Government, Harvard University. We are grateful to Alberto Abadie, Chris Avery, David Deming, Pascaline Dupas, Brian Jacob, Dale Jorgenson, Eliz- abeth King, Karthik Muralidharan, David McKenzie, Rohini Pande, Lant Pritchett, Jesse Rothstein, Douglas Staiger, Tara Vishwanath, and seminar participants at Harvard, NEUDC and BREAD for helpful comments on drafts of this paper. This research was funded by grants from the Poverty and Social Impact Analysis and Knowledge for Change Program Trust Funds and the South Asia region of the World Bank. The findings, interpretations, and conclusions expressed here are those of the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the governments they represent. † Earlier versions of this paper circulated under the title “Do value-added estimates add value? Accounting for learning dynamics”. 1