4. Characteristics of undercount Figure 1: Undercount by age and sex, raw PES data Figure 2: Undercount by population group and sex, raw PES data Figure 3: Undercount by marital status and sex, raw PES data
As noted earlier, there are a range of reasons people may be missed in the census. A consequence of this is that some segments of the population are more likely to be missed by a census than others. This section presents undercount results by various demographic characteristics to give an indication of the factors affecting undercount. The undercount rates in this section are calculated directly from the raw PES dataset. An alternative would have been to compare unweighted and weighted counts from the census but this would have revealed less detail about the separate characteristics as the XAID model selected and combined characteristics to create weighting classes. Thus, these rates do not directly reflect the adjustment factors as applied to the raw census counts. As the rates are included here to indicate broad trends rather than provide precise data, they are presented in graphs rather than tables. Figure 1, below, shows the undercount rate from the PES by age and sex. As has been observed in other countries, undercount is closely related to age and sex. A survey of undercount rates in Australia, Britain, Canada and the USA2 noted the following trends:
The authors of the report also noted that in Britain and the USA, women above 75 years of age were more difficult to enumerate than middle-aged women. Although Census ’96 had many unique problems and a higher undercount rate than the highly developed countries covered in the study, very similar trends were observed, as is clear in the following graph. Figure 1: Undercount by age and sex, raw PES data The groups most likely to be missed by the census are babies and young adult males. The most likely reasons for missing these groups are very different. Babies may be missed because the person completing the census questionnaire may not yet consider them as household members. Young adult males tend to lead more mobile lifestyles and be less attached to a particular household and thus are more difficult to enumerate. The decrease in the undercount rate for older age groups probably reflects the transition to more settled lifestyles. However, it is also interesting to note that, for most adult age groups, the undercount rate for males is higher than that for females. Figure 2 presents the undercount rate by population group and sex.
Figure 2: Undercount by population group and sex, raw PES data Africans and coloureds have higher undercount rates than whites and Indians. This is also reflected in the undercount rates by EA type as presented in Table 3.1. A large proportion of white and Indian people live in formal urban areas which are easier to enumerate, while a larger proportion of black and coloured people live in informal and non-urban areas where enumeration is more difficult. Regarding the sex breakdown, only among Indians is the undercount among women higher than among men – which, when one uncovers the finding, makes sense for likely cultural reasons. The link between lifestyle and likelihood of enumeration in the census is also shown in the undercount rates by marital status (Figure 3).
Figure 3: Undercount by marital status and sex, raw PES data Undercount is lowest for people who are married (civil or religious) and highest for people who are never married, living together or divorced/separated. These results are, in part, correlated to age and reflect the high undercount rates for young adults and the lower undercount rates for older adults. People who are divorced/separated are likely to be less attached to a household and this is reflected in the higher undercount rates observed. There are also appreciable variations by marital status in the differences between the undercount rates by sex. |
