| Abstract: | 戶籍資料是台灣地區大部分政府和學術機構從事抽樣調查工作引以為據的母體清冊。比起沒有建立全國人民登記制度的國家而言,這個母體清冊不但便利,而且節省了可觀的抽樣工作與經費。此外,戶籍資料也提供了許多與人口學和社會變遷有關的數據,對政府規劃部門和學術界的很多研究領域來說是非常重要的指標基礎。 本研究以民國七十九年台灣地區國民健康調查中的1031份家戶及4196份個人樣本為例,將樣本依照其原登錄地址是否與受訪當時有異,而區分成「原址樣本」與「遷移樣本」兩類,並分析「原址樣本」與「遷移樣本」兩類,並分析「原址樣本」與「遷移樣本」在基本人口、社會特性、就醫行為、健康行為上是否顯著差異。 研究結果顯示,在家戶樣本部分:遷移樣本在「居住地區」、「家庭結構型態」、「現址居住時間」、「訪談使用語言」等四項上,與原址樣本有顯著差異。在個人樣本部分:遷移樣本在「年齡」、「教育程度」、「有無職業」、「從業身分」、「保險類別」、「訪談使用語言」、「就醫行為」等五項人口社會特性上,與原址樣本有顯著差異。在就醫行為方面,遷移樣本的就醫方式,在「中藥房配藥」、「自行照顧」兩類上與原址樣本有顯著差異。因此,在台灣地區從事樣本調查工作,對於遷移樣本應儘量追蹤完訪,以避免忽略特殊樣本,提高樣本代表性。 其次,本研究再以1990年國民健康調查和1994年國民醫療保健調查的資料,參考學者專家的研究成果,檢驗了一部分戶籍登記項目的正確性,證實婚姻狀況有6.9%紀錄與受訪者自述不符,教育程度則有25.6%不正確。在戶籍登記人口與現住人口的差異方面,若以「人」為計算單位時有不到九成的人「籍在人在」,超過一成「籍在人不在」,超過一成「人在籍不在」。若以「戶」為計算單位,則只有大約五成五的戶現住人口和登記資料完全相符,其餘「籍多於人」和「人多於籍」者各約兩成,這些現象也因地區而異,因戶內人數的多寡而異。此外本文並以1990年內政部戶口普查的百分之五抽樣資料建立台灣地區戶內二十歲以上人口數的分布數據,證實如果等機率抽戶,再自戶中選擇一位受訪者,則理論上無法抽出一套有人口結構代表性的樣本。
The purpose of the study is to examine the differences between migrated samples and static samples on the demographic character, social economic status, medical behavior, and health habit. The data for the analysis came from the 1990 National Health Survey in Taiwan. Samples, 1031 households and 4196 persons, are classified into migrated samples and static samples according to whether the address the same as those were. There are two main findings in the study. Firstly, household migrated samples are significantly different from static samples on five variables which are resident area, family structure, resident time, and using language. Secondly, the individual migrated samples are significantly different from static samples on the following variables: age, education level, occupation, insurance, language, and medical behavior. Thus, the results of this study implicate the importance of following up migrated samples in the process of survey. The study also checks the accuracy of some items of the household registration records in Taiwan using data from the 1990 and the 1994 National Health Surveys and finds that the discrepancy rate between the marital status records and survey results is about 6.9%. There is also 25.6% inaccurate rate for the records of educational level. If counted by person, less than 90% of the people currently live in their registered households, over 10% of the people do not live in their registered households, and over 10% of the people live in places which they do not registered. If counted by household, only about 55% of the records are correct in both household registration and persons live in, other 45% are either greater or less than the registration record in terms of the number of people living in the household, or even the number is correct the persons` names do not match. This type of inaccuracy varies in different areas and also depends on the size of the household. The study also looks into the distribution of the household size using the 1990 census data and finds that it is not feasible to have a representative sample if households are selected first with equal probability and one adult respondent is then selected within the household due to the fact that a large portion of the younger people are clustered in bigger size households which account for only a much smaller portion of the total number of the households. |
