Variance Decomposition of Macro Level Data with Application to Unemployment in Northeast India

Analysis based on aggregated data is referred to as ecological analysis and the discrepancy arising when the analysis based on area level gives conclusion very different from expected relationship at unit level is termed as ecological fallacy. It is for this reason that census data are under-utilized for the exploration of cause and effect relationship. The uses of aggregate data from the census are confined to projection,estimation of rates, trend analysis and compositional changes. The three sources of ecologic fallacy pertain to situations where there is some form of group effect. This includes situations where there is a failure to distinguish constructs at different levels (e.g., mean group X is assumed to measure the same thing and individual-level X), where something about the groups is associated with individual-level predictors of the outcomes (mean group X is associated with other individual-level factors related to Y), or where some social process results in the grouping of persons by the dependent variable. This paper describes statistical technique which can be adopted for macro level analysis by decomposing variance under multilevel regression model.