Small domain statistics: a review

Local data are required at short intervals of time mainly for undertaking regional development projects on ascertaining area‐wise socio‐economic requirements, proper devolution of funds and resources from central pools and allocation of legislative seats. A census is an infrequent source for them and is also incomprehensive. Official records and registers are a second source but are generally inadequate. Sample surveys are a recognized third alternative. But even for large total sample sizes, when numerous local areas are involved, many are not represented well enough in samples. Consequently, for totals of small areas classical survey estimates turn out inefficient having large mean square errors admitting uselessly wide confidence intervals. Improved estimation procedures however are rapidly emerging based on postulated models which encourage borrowing strength from samples outside local areas and from past data. A brief review is attempted illustrating applications.