Estimating the critical and sensitive periods of investment in early childhood: a methodological note.

The identification of critical periods in early human development requires statistical analyses beyond simple cross tabulation of correlations of observed variables. This paper provides an overview of different quantitative methods available for the statistical analysis of longitudinal data regarding child development, and in particular the identification of critical and sensitive periods for later abilities. It draws heavily on the work on human skill formation developed by the economist James Heckman, which treats ability as a latent variable and explains its formation through the simultaneous estimation of structural equations of investments and achieved abilities across time. We distinguish between two specifications of the ability formation function. One of them (the 'recursive') format explains current ability as a function of the ability and investment at the immediately preceding period. The other (the 'non-recursive') format explains current ability as a function of a series of past investments. In order to fully examine critical and sensitive periods of investments, the non-recursive formulation needs to be used. Furthermore, true abilities of an individual cannot be directly observed: what we observe are the test scores, for example, on reading and writing. We outline an approach, structural equation modelling, that treats actual test scores as measurements of the latent ability variable, and show how it can be used in the recursive and non-recursive formulation. In order to fully examine critical and sensitive periods of investments, we argue that the non-recursive formulation of this structural model is necessary. However, the non-recursive formulation requires more data than the recursive formulation, and to the best of our knowledge, has never been used in the identification of critical and sensitive periods in early childhood development.