Interviewer instructions
This weight factor compensates for various features of sampling in givein countries, to make the samples replicate the national population parameters more closely. For example, the 1981 surveys in Western Europe, the United States, Canada and Mexico oversampled (by approximately 50 percent) the youngest group aged 16-24. These respondents receive proportionately less weight in this variable. The samples from China, India, Nigeria undersample the illiterate and rural portions of the public and oversample the more educated and urban portions; the weight variable is designed to correct for this problem by giving greater weight to the less educated. Both the 1981 and 1990 South
African samples were stratified by race, interviewing
approximately as many whites as Blacks; the weight variable
corrects for this. This variable also corrects for obvious deviations from national population parameters in age and education in other countries. In most cases, the more highly educated are oversampled, and are accordingly weighted less heavily than the less educated. In the 1990 Italian sample, however, the more educated are substantially undersampled, and are weighted more heavily to compensate for it.
Finally, the 1990 Spanish sample has a much larger N than most other samples, which would give it disproportionate importance in any analysis involving pooled samples; it is down-weighted. Similarly, this study includes many small countries, and their combined Ns would far outweigh the results from the larger countries: unweighted, the Nordic countries plus the Baltic countries would outweigh India, China, the United States and Russia. This weight factor gives greater weight to the more populous countries than to the less populous ones, so that pooled analyses (which are often convenient) more closely approximate global reality. The weighted N of the combined 67 surveys assembled here is 89,672, as compared with the unweighted N of 89,909.
