A Comparative Study of the Couple in the Social Organization of Sexuality in France and the United States - Statistical Data Included

Journal of Sex Research, Feb, 2001 by John H. Gagnon, Alain Giami, Stuart Michaels, Patrick de Colomby

Sample Selection

In any comparative research using independently collected data sets there are always problems of comparability of methods (Spira, Bajos, Giami, & Michaels, 1998). In France, the sample was drawn from a list of telephone subscribers, which is a kind of sample of households. In order to arrive at a sample of individuals, a single adult in each household was randomly selected to answer the questionnaire (on the telephone). Prior to the ACSF, a pilot survey comparing telephone interviews with face-to-face interviews was carried out. No significant differences in the responses obtained by these two methods were found (ACSF Investigators, 1992). In addition, after about 30 mostly sociodemographic questions, a filter card was used to direct respondents to either a short or a long form of the questionnaire. The French filter card was:

   Do you fall into one of these categories, without telling me which one: You
   were born the 4th, the 17th or the 20th of any month. You have had sexual
   intercourse with at least two different people in the past 12 months. In
   the past 5 years, you have had sexual intercourse at least once with a
   person of the same sex as you. In the past 5 years, you have paid to have
   sexual intercourse at least once. In the past 12 months, you have used a
   soft drug (hashish, marijuana, etc.) or a hard drug (cocaine, heroin,
   etc.). You are a hemophiliac (Spira & al. 1994, p. 28).

The purpose of the filter was to produce a larger sample of persons considered to be at risk while maintaining a control group of persons not necessarily at risk. The short form was administered to 15,235 persons and the longer one (that also included all of the questions on the shorter form) to 4,820 persons.

The U.S. sample was also based on a sample of households and, in addition, included an oversample of Blacks and Hispanics. The U.S. survey was carried out face-to-face (including some self-administered forms), usually in the home of the selected respondent. The U.S. sample was of English speakers ages 18 to 59 living in households. All respondents received the same questionnaire.

Representative samples of the population require that each respondent have a known probability of being selected. Samples such as those used in the French and U.S. surveys are produced using complex sample designs, which result in unequal probabilities of selection. Analyses of such a sample require weighting cases to produce a posteriori the equivalent of equiprobability, which allows one to produce unbiased estimates of parameters describing the population from which the sample was drawn. Unequal probabilities of selections arise for a number of reasons. In both the French and the U.S. surveys only a single eligible member of each household was interviewed among those living there. In the French survey, persons at risk were oversampled. In the U.S. survey, Blacks and Hispanics were oversampled. All analyses use weighted cases to produce correct proportions and percentages of the population as a whole. In addition, given the complex sample designs used in each case, special procedures were necessary to compute standard errors and other components of inferential statistics.(2)