University of California, Los Angeles, USA.
Demography. 1967 Jun;4(2):759-72. doi: 10.2307/2060315.
This article presents estimates of the sources and the extent of observation errors in different questionnaires and methods used to collect birth and death data in the 1961-63 multi-purpose sample survey of Morocco.The questionnaires used in the analysis of the three survey rounds were a list of household members (Rounds1 and 2) and a roll-call (Round3); retrospective death (Rounds1, 2, and 3) and birth (Round 3) queries; a date-of-birth tabulation (Round 2); and a household check-sheet to explain differences between Rounds 1 and 2. All available questionnaires for a given household were brought together and collated to provide several sources of information on births and deaths and a basis for assessing errors.From this analysis, the survey attempted to define the nature and to estimate the frequency of the errors which would have occurred if more restricted types of survey design had been used. Results, based on the period between Rounds 1 and 2, led to three major conclusions.First, if vital data had been collected with a single-round retrospective procedure, gross error (over enumeration plus underenumeration) would have been 17 percent for births and 36 percent for deaths. There is a net error of overenumeration of 3 percent for births (1.4 per1,000population) and 9 percent for deaths (2.3 per1,000population).Second, if two rounds were available to permit a combination of household composition follow-up and a retrospective mortality questionnaire, overenumeration would be almost entirely eliminated and underenumeration would be noticeably reduced. Third, most of the remaining errors of underestimation may be attributed to (1) an estimated number of infants born and deceased between two rounds and missed by all questionnaires, (2) matching failures caused by the absence of adults at Round 1, and (3) matching errors.
本文估计了在摩洛哥 1961-63 年多用途抽样调查中用于收集出生和死亡数据的不同问卷和方法的来源和观测误差的程度。分析中使用的问卷包括家庭成员名单（第 1 轮和第 2 轮）和点名（第 3 轮）；回顾性死亡（第 1、2 和 3 轮）和出生（第 3 轮）查询；出生日期列表（第 2 轮）；以及用于解释第 1 轮和第 2 轮之间差异的家庭检查表。将一个家庭的所有可用问卷汇集在一起并进行核对，以提供关于出生和死亡的多个信息来源，并为评估误差提供依据。从这项分析中，调查试图定义如果采用更严格的调查设计类型会出现的错误的性质和频率。基于第 1 轮和第 2 轮之间的时期的结果，得出了三个主要结论。首先，如果使用单轮回顾性程序收集生命数据，那么出生时的总误差（多报加漏报）将为 17%，死亡时的总误差为 36%。出生时的净多报误差为 3%（每 1000 人中有 3 人），死亡时的净多报误差为 9%（每 1000 人中有 23 人）。其次，如果有两轮可供使用，允许家庭构成的后续调查和回顾性死亡率问卷相结合，那么多报将几乎完全消除，漏报将明显减少。第三，大部分剩余的低估错误可能归因于（1）两轮之间估计的婴儿出生和死亡人数，所有问卷都漏报了；（2）第 1 轮时没有成年人而导致的匹配失败；（3）匹配错误。
