Coverage

The term coverage, as used in survey research, indicates how well the sampling units included in a particular sampling frame account for a survey's denned target population. If a sampling frame does not contain all the units in the target population, then there is undercoverage of the population. If the frame contains duplicate units or other units beyond those contained in the population, then there is overcoverage. Undercover-age and overcoverage do not necessarily mean there will be coverage error associated with the frame.

Overcoverage occurs when members of the survey population are erroneously included in the survey sampling frame more than once or are included erroneously. Noncoverage (including undercoverage) occurs when members of the targeted population are erroneously excluded from the survey sampling frame. The meaning of the term noncoverage is not the same as the meaning of unit nonresponse, which is the failure to obtain complete survey data because of issues such as noncontacts, refusals, lost questionnaires, and so on.

Both overcoverage and noncoverage can occur at several junctures during the survey process. For example, in population surveys in which the sample is selected in two or more stages to obtain estimates of persons within households, coverage errors may occur at any or all stages when creating the sampling frame of primary sampling units, during field listing of housing units, or when creating a household roster of persons within a given family. Noncoverage that occurs during field listing can result if members of the survey sample are excessively expensive to locate or are part [Page 160]of multi-unit structures, or if maps do not accurately display the sampling area. Survey coverage is affected by the amount of time that has lapsed between obtaining the information for constructing the frame, creating the frame, drawing the sample, and finally collecting the data by methods such as personal visit, telephone, mail, Web, or by abstracting records. Several months or years may have passed during this time period, and many changes may have occurred to the units in the initial sampling frame that will not be reflected in the final sample.