Population Coverage and Undercount
Although the ideal of census agencies is a 100% enumeration, most censuses fall short of that, and the rate of response is partly dependent on whether a census is compulsory or not. Although participation in the census in most countries is in theory compulsory, in reality it is not always possible to enforce. In the United States census of 2000, which relied more on promotion than compulsion, the final response rate was 67%, slightly higher than the 65% achieved ten years earlier. Even in countries where compulsory participation is more rigorously enforced, some individuals will be missed, and a few mobile individuals may be counted more than once, so that the net effect is a population undercount.
A post enumeration survey (PES) is the most common way of estimating the accuracy of census coverage not only in terms of population undercount, but also in relation to particular characteristics of the population. A PES is based on intensive sampling of particular populations and areas usually soon after the census itself to estimate the proportion of population missed by census enumeration. Target groups may include the homeless, transients, and people absent from their usual residence at census moment. Target areas may include areas with relatively lower educational levels, areas with higher numbers of undocumented migrants, and remote areas. There is a significant politics surrounding population coverage and undercount. Since undercounts are more likely in areas of lower socioeconomic status, if these are not rectified by a PES or similar instrument, there may be a number of disadvantages for such areas. Since most electoral systems are based on proportional representation, such an area will be underrepresented in the political system. Further, in countries such as the United States where public funding is allocated according to population, health, education, and other social funding from federal and state sources will be reduced if the undercount is not considered.
Sources of Census Error
Data errors may occur when a respondent has provided incorrect or ambiguous information or when data are being coded and entered. Respondent errors fall into several categories:
- accidental inaccuracies, for example, visitors overlooked, babies forgotten, and questions misunderstood;
- intentional misrepresentations, for example, hiding of undocumented migrants and misreporting of income; and
- variables with inherent ambiguity, for example, ethnicity (see below).
Errors related to coding and entering of data may be simply technical mistakes, but they also may relate to problems with classification of particular variables, and examples discussed below include classification of households and families, and classification of ethnicity and religion.