An Analysis of Sample Attrition in the PSES Panel Data

Panel household surveys study the dynamics for individual households, including their well-being, child schooling, and labour market dynamics, and they can also be used to observe who benefits or loses from general economic development. The MIMAP panel survey, covering 300 households in 60 districts, is the largest panel survey carried out in Pakistan. This paper addresses two very legitimate concerns regarding the panel data? the extent of attrition and the degree to which attrition is non-random. The paper describes in detail the factors responsible for attrition in this panel at the provincial as well as the urban-rural levels. The paper first examines the extent of attrition and then differentiates the characteristics of the attriting household from the non-attriting household, and how this attrition affects the coefficients of standard background variables. It shows that attrition was higher among the poorest households as measured by the per capita expenditures. However, no significant difference could be found between the set of coefficients for attritors versus non-attritors for indicators of interest, i.e. consumption and poverty. This shows that coefficient estimates of standard background variables are not affected by sample attrition, implying that an attrition of more than 22 percent sample of PSES is not a serious problem for obtaining consistent estimates. This analysis of the extent and nature of PSES panel attrition suggests that researchers can use this panel data set with confidence to explore the dynamic relationship in Pakistan. It also indicates clearly that it is worth investing in another round of the PSES panel.