The current study employs definitions of school outcomes similar to those used by
World Bank (2004) and it estimates multivariate probits that are similarly specified.
However, we use the NFHS data rather than the NSS, which is useful in that it
provides an opportunity to cross-check the results of one study against the other. A
contribution of the current study is that it uses repeated cross-sectional data (two
rounds of the NFHS) to investigate the growth in schooling indictors. It then assesses
the extent to which (a) the predictor variables actually change and (b) the parameters
are stable over time. We find that the predictor variables change much less than
hypothesized in the illustrative simulations conducted in World Bank (2004).3 For
example, in scenario-3, the assumed annual change in male and female years of
schooling is 0.25 and 0.3 respectively. Between 1992/3 and 1998/9, these variables
increased (for the sample of 6-11 year olds) by only 0.055 and 0.066 years per year
respectively. We also find that the parameters are not stable over time, which makes it
very difficult to extrapolate to the future. Indeed, we find that almost all of the growth
in schooling can be attributed to changes in the elasticities. We caution against the
common practice of making predictions on the assumption of stable parameters, while
recognizing that there may be no better alternative. We conclude that the prospect of
India attaining universal primary attendance is good, but that the prospect of attaining
universal completion rates in primary school is bleak unless a major intervention is
undertaken.