Main Sample
When initiating the PSFD project, the main sample of the Taiwan PSFD survey was set to the adult population aged between 25 and 64. Due to cost concerns, the first-wave surveys for the main sample (Questionnaire RI) who met the age requirement were conducted in 1999, 2000, and 2003. In 1999, the first-wave survey for the main sample who were born in 1953-64 was conducted by face-to-face interview. The surveys for the main sample who were born in 1935-54 and 1964-76 were conducted in 2000 and 2003. Since then, the refreshment sample of the younger cohorts has been added to the main sample every six to eight years. The first-wave surveys for the main sample who were born in 1977-83 and 1984-91 were conducted in 2009 and 2016 respectively.
The sampling approach adopted in the first-wave surveys for these five groups of the main sample was the stratified three-stage PPS (probability proportional to size) sampling method, with the sampling frame being registration records covering Taiwanese citizens who fit the birth-year requirement. The townships and districts of Taiwan are first stratified by their degree of urbanization. In the first stage of sampling, townships and districts are randomly drawn from each stratum. In the second stage, villages are randomly drawn from each selected township or district. In the third stage, individuals who fit the birth year criterion are randomly drawn from each selected village.
The sampling for the first-wave surveys of the five groups of the main sample was slightly different in two respects. First, the stratification of townships and districts used in the 1999, 2000, and 2003 first-wave surveys was different from that of the 2009 and 2016 first-wave surveys. For the first-wave surveys conducted in 1999, 2000 and 2003, the stratification of townships and districts was based on Luo Qi-Hong’s (1990) study entitled ‘Types of developments of villages and townships in Taiwan’ (Taiwan Economy Monthly, 190, 41-68), while the strata for the first-wave surveys conducted in 2009 and 2016 were from Pei-Chun Hou et al.’s (2008) study entitled ‘The typology of townships in Taiwan’ (Survey Research – Method and Application, 23, 7-32). The other difference in the sampling procedures among these first-wave surveys was the strategy used to cope with anticipated non-response of selected individuals. In the 1999 and 2000 first-wave surveys, three samples were drawn from the sampling frame, with the sample size of the first two being the targeted sample size, and the sample size of the third being half of the full targeted sample size. During the fieldwork, the randomly drawn individuals were assigned to the interviewers on a set-by-set basis. As to the 2003, 2009, and 2016 first-wave surveys, sample size was inflated, with the number of randomly drawn individuals ranging from 1.6 to 2.1 times the targeted sample size. The interviewers attempted to contact all the assigned individuals during the fieldwork.
Child Sample
The sample of the PSFD project has been extended to the children of the main respondents since 2000. The children born after 1977, once they reach age 16 at the time of the survey, are included as part of the child sample and re-interviewed biennially. When the second-generation respondents turn 25, they are treated as main respondents. As long as they finish the first-wave questionnaire of the main sample, they are re-interviewed along with the original main respondents using the same questionnaire. Unlike the original main respondents, the 2nd-generation respondents are not randomly selected. The demographic information for the children is retrieved from the questionnaire of the original main respondents. For the children who are under 20, in addition to their own consent, consent from the original main respondents is required for the interview of the child (using Questionnaire C). The children who are not successfully contacted and interviewed at the age of 16 are approached again in future surveys.
As indicated above, the sample of the PSFD survey is composed of the main respondents and their children who meet the age and birth year criteria. Up till now, there has been no attempt to include the children of the child sample (i.e. 3rd generation) as part of the sample.