Data Collection Procedures

The NUJLSOA sample was selected using a multistage stratified sampling method. The sampling was done by, first, stratifying the 47 prefectures throughout Japan into 11 regions. Approximately 3000 municipalities were then stratified by population size within regions and prefecture and a systematic sampling method was used to select 340 primary sampling units. There are three sources through which a national list sample of individuals can be selected from primary sampling units in Japan: the National Residents Registry System, the list of eligible voters, a listing of all households developed by Central Research Services Incorporated. Registration of persons in Japan is considered to be universal and accurate because it is a legal requirement to report any moves to local authorities within two weeks. The National Registry List is the most desirable source of survey respondents because it is viewed as the most up to date. The eligible voter list, while based on the Registry, may be less up to date depending on the most recent date of updating of this list from the National Registry List. The reason that any source other than the National Residents Registry is used to identify potential respondents is that some municipalities will not allow access to this list even for research purposes. In this case the list of voters was used. If a municipality would not allow access to either list, the list of housing units developed by Central Research Services Incorporated was used. This list is regarded as relatively accurate as it has been being maintained for almost 50 years and is regularly used for national samples. The residents chosen for the NUJLSOA in 298 primary sampling units were selected based on the list from the National Residents Registry of Japan. In another 41 units, sample members were drawn from the list of eligible voters. In 1 unit, sample members were drawn from the Central Research Services Inc. housing unit list. This required identifying older persons in households chosen from a map of the area. From each of the 340 areas 6-11 persons ages 65-74 were selected and 8-12 persons at ages 75 and over were selected for the sample. The population 75 years and older was oversampled by a factor of 2. Weights have been developed for respondents to the first wave of the survey to reflect sampling probabilities. Weights for the second wave are under development. To develop the refreshed sample at wave II an estimate was made of the size of the 65 and 66 year old population and a target list of potential sample members was drawn assuming a 65 percent response rate.

For a description in Japanese, see http://www.nihon-u.ac.jp/cin/nujlsoa/

The interviews were performed using structured questionnaires administered by a private survey firm, the Central Research Services Inc (CRS). This is the most experienced firm doing longitudinal surveys in Japan. The CRS was organized in 1954 through the unification of the National Public Opinion Research Institute and the research division of the Jiji Press, Ltd. The CRS has permanent facilities at its Tokyo headquarters and four regional offices in Sapporo, Nagoya, Osaka and Hiroshima, and in the 52 branch offices throughout Japan. They offer full services for questionnaire design, sample design, editing, and coding. The CRS has over 600 trained interviewers who do more than 250 surveys with over 300,000 interviews a year. This firm employed a group of 331 interviewers who were trained by the CRS staff under the direction of Dr. Saito to administer the NUJLSOA questionnaire. An attempt to encourage participation in many communities was made by providing information on the forthcoming survey to local newspapers. Newspapers articles presented the forthcoming survey as an important undertaking of a major University research group whose findings should be relevant to older peoples’ lives. Respondents were given a small gift worth about $10 US to express gratitude for their participation. Continued cooperation of actual respondents to the survey has been encouraged by regularly communicating with them. They have been sent 2 New Year’s cards, 2 other greeting cards, and some summary results from the first interview by CRS INC.

As noted above, the questionnaire was designed to be similar to that of surveys in other countries as well as to be appropriate for Japanese elders. When instruments needed translation from English, they were translated into Japanese, and then back-translated into English. Attention was paid to having both an accurate translation and appropriate Japanese wording. Standard instruments such as the CES-D were already available in Japanese. Questions on functioning and health states had been asked in Japanese surveys previously but an effort was made to have many of these questions be identical to the format used in the LSOA and AHEAD so comparative work would be possible.

There were two pretests before the first wave of the survey, each consisting of 50 respondents. Pretest respondents were a convenience sample from the Tokyo metropolitan area. In addition, two residents of a hospital were pretested. There was one pretest before the second wave of the survey consisting of 50 respondents. These pretests were used to refine the instrument, adjust the flow of the questions, and improve interviewer training. The questionnaire was designed to be answered by the respondent where possible and a proxy when necessary. Proxies were used when the respondent was unable to answer because of physical or mental health problems. A respondent was required to grant permission for a proxy to answer the survey. At the first interview, 12 percent of responses were provided by proxies.

The original sample list for the first interview consisted of 6,700 potential respondents selected in October, 1999: 2,922 potential respondents ages 65-74 and 3,778 respondents ages 75 and above. Knowledge of past surveys in Japan and budget constraints were used to select 6,700 as the number of potential respondents. The original target was a 75% response rate and 5000 eventual respondents. Of the original list of 6,700 individuals, 4,640 (69.3%) responded at the initial wave of data collection in November and early December 1999. Three months later, in March 2000, an additional attempt was made to interview 710 sample persons who were either absent, institutionalized, or who did not strongly refuse to participate at the first interview. This resulted in an additional 357 respondents bringing the total sample to 4,997 and an overall sample response rate of 74.6% which came very close to meeting the initial target. Of the 357 respondents added in the March follow-up, 49 were institutionalized during the November interview. In order to reduce their burden, these institutionalized respondents were given a shorter form of the interview in March. The second wave of data was collected in November 2001. The third wave was collected in November 2003 will be released at a later date.