The Rational Evaluation of Official Statistics
IN contrast to the position examined in Chapter 2, the more orthodox texts on social statistics assume that, in principle, a true allocation of individuals into response categories is obtainable. Any given statistical distribution is to be evaluated with respect to its deviation from the presumed true distribution, that is, with respect to its net or gross response error. Survey methodology aims to obtain the minimum possible error within the limits set by time and available resources. The presumed true distribution is assumed to be independent of any investigation. I shall argue that such a conception of error is unacceptable as far as empirically given populations are concerned and that, in fact, the plausibility of this conception depends on ignoring the theoretical character of survey or census categories and of the instructions which determine how individual cases are to be assigned to them. These points are argued with the help of an extended example dealing with agrarian statistics.
KeywordsIncome Coherence Expense Sorting Folk
Unable to display preview. Download preview PDF.