Inter-Censal Population Survey 2004 - Cambodia

Abstract
---------------------------

The Cambodia Inter-Censal Population Survey, 2004 was designed not only to obtain the much-needed demographic data following the census, but also to serve as a means to train the staff of the NIS and Provincial Planning Offices in demographic data collection.

There are plans to produce in-depth studies on fertility, mortality, migration, literacy and education, labour force, housing and household amenities, and population projections based on the results of the survey.

The Cambodia Inter-Censal Population Survey 2004 (CIPS) is a nationally representative sample survey taken between two censuses, the 1998 census and the proposed 2008 census, in order to update information on population size and growth and other population characteristics as well as household facilities and amenities. Due to the national elections and administrative issues, the CIPS was undertaken in March 2004 instead of 2003, which would have been the five-year midpoint between the 1998 and 2008 censuses.

The conduct of the CIPS 2004 is an important step in the creation of a continuous flow of data that will allow Cambodia to prepare plans and programmes supported by a strong database.

The Cambodia Inter-Censal Population Survey 2004 was conducted with the objective of providing information on the following indicators:
- Sex, age and marital status
- Births and Deaths
- Migration status
- Literacy/Educational level
- Economic characteristics
- Housing and household amenities
- Other population and household information

These fresh data will allow for calculations and reliable projections of:
- Population size and growth
- Fertility
- Mortality
- Migration

The survey was also intended to train the national staff in sampling, data collection, data processing, analysis and dissemination.

Geographic coverage
---------------------------

National

Analysis unit
---------------------------

Individual, Household

Universe
---------------------------

All Population and housing for all regular households in Cambodia excluding special settlements and institutional households.

Kind of data
---------------------------

Sample survey data [ssd]

Sampling procedure
---------------------------

The sampling design for the CIPS 2004 is a three-stage stratified cluster sampling design, it is a probability sample selection of 100 percent of the Cambodian villages coverage areas, the survey covered only regular households and excludes special settlements and institutional households.

The CIPS 2004 was conducted in a nationwide representative sample of 21,000 households within selected 700 villages (primary sampling units) out of 13,886 villages in Cambodia. The 700 villages were selected from updated frame (list of villages for Cambodia).

The General Population Census 1998 databases of the National Institute of Statistics together with the new updated list of villages that were excluded in the general population census of 1998 was used as the sampling frame for the sampling design of the CIPS 2004.

The frame has the following identification particulars:
1- Province code
2- Province name
3- District code
4- District name
5- Commune code
6- Commune name
7- Village Code
8- Village name
9- Size of village (number of households)
10- Area code (1 = Urban, 2 = Rural)

A three-stage sample design has been used for the CIPS. In the first stage a sample of villages was selected. The villages were implicitly stratified into 45 strata (21 provinces each with rural/urban strata i.e. 42 strata plus 3 provinces each totally urban, i.e. 3 urban strata). The villages were selected using linear systematic sampling with probabilities proportionate to size (PPS). The size measure used for the selection was number of households in the village according to the 1998 Census with estimation for a few additional villages not in the 1998 census frame.

In the second stage one Census Enumeration Area was selected randomly (in the head office) in each selected PSU. At the beginning of the fieldwork all households in the EA were listed. A systematic sample of 30 non-vacant households was selected as the third stage of selection.

The listing of households in the EA would become cumbersome if there are many households in the EA. This might be the case when the enumeration area had grown substantially since the census. When the EA was large (population wise) the interviewer was instructed to split the EA into two or more approximately equal-sized segments and to select one segment randomly. All households in the selected segment were listed. Out of the 700 Sample PSUs, 598 were from the rural super stratum and the remaining 102 were from the urban super stratum. For more information on sampling for the survey the general report at national level may be referred to.

Note: All provincial headquarters were treated as urban. In the case of Sihanoukville, Kep and Pailin, the entire province was treated as urban. In Phnom Penh province, the four districts of Doun Penh, Chamkar Mon, 7 Makara and Tuol Kouk were classified as urban. All the remaining areas of the country were rural. Further, urban and rural areas are being reclassified in Cambodia. While these reclassifications have already been drafted, they have not yet been approved by the Royal Government of Cambodia. Upon endorsement and adoption, the new classifications will be used in future census/surveys.

Mode of data collection
---------------------------

Face-to-face [f2f]

Research instrument
---------------------------

The draft questionnaires for the CIPS 2004 were more or less on the 1998 General Census pattern. Some modifications, however, were made by adding new questions on

(i) Whether children aged 0-14 living with own mother
(ii) Whether a person's mother is alive and
(iii) Details of deaths in households in the last one year with focus on maternal mortality.

Questions mentioned at (i) and (ii) were intended respectively to estimate fertility (by application of own child method) and mortality (by application of orphan hood method). The questions to be included were carefully considered by a Working Group of Cambodia Inter-Censal Population Survey 2004, whose members were mostly from Ministries, NGOs and International Agencies. The Questionnaires were tested twice in the field (both urban and rural) by NIS staff in November 2003. The purpose of the pre-test was to have a full-dressed rehearsal of the whole process and particularly to test the questions in the field so as to make corrections in wording or definitions and to estimate the time taken for enumeration area mapping, house listing, sampling and enumeration of selected household. Based on the pre-test experience the questionnaires were modified and finalized.

Two types of questionnaires were used in the CIPS 2004: Form A House-list and Form B Household Questionnaire.

The Form A was used to collect information on buildings containing one or more households during the preliminary round preceding survey night (March 3, 2004). The information collected related to: construction material of wall, roof and floor, whether it is a wholly or partly residential building, number of households within the building, name and sex of head of household and number of persons usually living in the household.

The Form B, which has five parts, was used for survey enumeration in the period closely following the reference time.

In Part I, information on usual members of the selected household present on survey night, visitors present as well as usual members absent on survey night, was collected.

Part II was used to collect information on each usual member of the household and each visitor present on survey night. The information collected included: full name, relationship to household head, sex, age, natural mother, child aged 0-14 living with own mother, marital status, age at first marriage, mother tongue, religion, place of birth, previous residence, duration of stay, reason for migration, literacy, full time education and economic characteristics.

Part III was used to collect information on females of reproductive age (15-49) as well as children born to these women.

The information collected in part IV related to household conditions and facilities: main source of light, main cooking fuel used, whether toilet facility is available, main source of drinking water and number of living rooms occupied by household.

Part V was used to record the following information in respect of deaths in the household within the last one year:- name of deceased, sex, relationship to head of household, age at death, whether the death has been registered with the civil authorities or not, the cause of death and maternal mortality information.

Cleaning operations
---------------------------

The completed records (Forms A, Form B, Form I, Form II, Map, and other Forms) were systematically collected from the provinces by NIS Survey Coordinators on the due date and submitted to the team receptionist at NIS. NIS Survey Coordinators formed into three teams of two persons were trained during March 7-10 to receive and arrange the completed forms and maps for processing after due checking form the field. Control forms were prescribed by DUC to record every form without any omission. These records were carefully checked, registered and stored in the record room. Editing and coding of the questionnaires were done manually, after which the questionnaires were submitted to the computer section for further processing. The instruction for editing and coding were revised and expanded. Training on editing and coding was conducted for senior staff, who in turn had to train other editors and coders.

The purpose of the editing process was to remove matters of obvious inconsistency, incorrectness and incompleteness, and to improve the quality of data collected. Coding had to be done very carefully in respect of birthplace and previous place of residence by using the district and province codes, and occupation and industry by using the UN International Standard Classification of Occupation (ISCO) and the International Standard Industrial Classification (ISIC) respectively. For these purposes, NIS utilized staff with sound knowledge and experience of the survey and its concepts. Those who worked as trainers or supervisors were put on this job supplemented by well-trained and tested staff. Editing and Coding was done by two teams (each with six editors and one team leader); so that one of the editors who were trained specifically in occupation/industry coding should do the coding for columns 20 and 22 of part 2 household questionnaire. The work of team members was completely checked by the Team leaders. The training on editing and coding was done from 23 to 26 March. The manual processing commenced on March 29 and was completely done by the end of May 2004.

Response rate
---------------------------

Response rate is 95 per cent.

Sampling error estimates
---------------------------

Calculations of sampling errors have been made for some estimates of totals, means and proportions for variables in Form B.

The software used for the calculations is STATA 8.0. For the calculations presented here we have assumed that stratification was done on provinces and urban/rural (an implicit57 stratification on province and urban/rural was used for the sample selection).

In seven of the 45 strata there are only one PSU (EA) selected. This causes a problem for the standard error calculations, it is not possible get standard errors in these strata. In these strata we have split the sole EA in two parts and defined the parts as two PSUs.

The standard errors are generally rather small for estimates for major domains like urban/rural and men/women. The coefficients of variation (CV) are below 1% in many cases. The coefficients of variation are substantially higher for provincial estimates, especially for provinces with a small sample (e.g. province19). Design effects (Deff) have been calculated for some estimates. They are, as expected, quite low for estimates of demographic characteristics. They are considerably higher for estimates of socio-economic characteristics like employment status (also as expected). For the demographic characteristics "age at first marriage" and "marital status" we find design effects below 5 for major domains like men/women and urban/rural. The socio-economic characteristics are typically more "clustered" than the demographic characteristics, this shows up in generally higher design effects. For the major domain estimates we find design effects up to 20 and occasionally very high values of 200 or more. These "freak" values occur when the sample in terms of number of PSUs is small and when the PSU averages (or proportions) show large variation. One example is the design effect of 285 for the estimate of proportion of government employees in urban areas. The proportion is varying substantially between the 102 PSUs in the domain, the range is from 0 % to75%.

摘要
---------------------------
柬埔寨2004年间的普查人口调查旨在不仅获取普查后迫切需要的统计数据，而且作为培训国家统计局和省级规划办公室人员在人口数据收集方面技能的手段。

计划基于调查结果进行深入的生育、死亡率、迁移、识字率和教育水平、劳动力、住房和家庭设施以及基于调查结果的人口预测研究。

柬埔寨2004年间的普查人口调查（CIPS）是在两次普查之间进行的全国代表性样本调查，即1998年普查和拟议的2008年普查之间，旨在更新关于人口规模和增长以及其他人口特征以及家庭设施和便利设施的信息。由于国家选举和行政问题，CIPS于2004年3月进行，而不是原定的2003年，后者是1998年和2008年普查之间的五年中点。

CIPS 2004的开展是创建连续数据流的重要一步，这将使柬埔寨能够制定基于强大数据库的计划和项目。

柬埔寨2004年间的普查人口调查旨在提供以下指标的信息：
- 性别、年龄和婚姻状况
- 出生和死亡
- 迁移状况
- 识字率/教育水平
- 经济特征
- 住房和家庭设施
- 其他人口和家庭信息

这些新数据将允许进行以下指标的计算和可靠的预测：
- 人口规模和增长
- 生育率
- 死亡率
- 迁移

调查还旨在培训国家工作人员在抽样、数据收集、数据处理、分析和传播方面的技能。
地理覆盖范围
---------------------------
全国
分析单位
---------------------------
个人、家庭
总体
---------------------------
柬埔寨所有常规家庭的人口和住房，不包括特殊定居点和机构家庭。
数据类型
---------------------------
样本调查数据 [ssd]
抽样程序
---------------------------
CIPS 2004的抽样设计是一个三阶段分层聚类抽样设计，它是对柬埔寨100%村庄覆盖范围的概率样本选择，调查仅覆盖常规家庭，不包括特殊定居点和机构家庭。

CIPS 2004在全国范围内进行的代表性样本调查中，选择了700个村庄（初级抽样单位）中的21,000户家庭，这些村庄是从13,886个村庄中选出的。700个村庄是从更新的框架（柬埔寨村庄名单）中选出的。

国家统计局1998年总人口普查数据库以及新更新的1998年总人口普查中排除的村庄名单被用作CIPS 2004抽样设计的抽样框架。

框架具有以下识别细节：
1- 省代码
2- 省名
3- 区代码
4- 区名
5- 县代码
6- 县名
7- 村代码
8- 村名
9- 村庄规模（家庭数量）
10- 区域代码（1 = 城市，2 = 农村）

CIPS已使用三阶段样本设计。在第一阶段，选择了村庄样本。村庄被隐式分层为45个层（每个省份21个，即农村/城市层，即42个层加上3个完全城市省份，即3个城市层）。使用与规模成比例的概率（PPS）进行线性系统抽样来选择村庄。用于选择的规模衡量标准是根据1998年普查的家庭数量，并对一些不在1998年普查框架中的额外村庄进行了估计。

在第二阶段，在每个选定的初级抽样单位中随机选择一个普查登记区域（在总部）。在实地工作开始时，列出了EA中的所有家庭。在第三阶段，选择了30户非空户作为系统样本。

如果EA中有许多家庭，那么在EA中列出家庭可能会变得繁琐。当EA在普查后大幅增长时，可能会出现这种情况。当EA在人口上很大时，调查员被指示将EA分成两个或更多大致相等的部分，并随机选择一个部分。列出所选部分中的所有家庭。在700个样本PSU中，598个来自农村超级层，其余102个来自城市超级层。有关调查的抽样信息，可以参考国家层面的总体报告。

注意：所有省级总部都被视为城市。在磅逊、吉普和帕艾林的情况下，整个省份都被视为城市。在金边省，四个区（斗蓬区、查卡尔蒙区、7号马卡拉区和陶克区）被归类为城市。该国的其余地区均为农村。此外，柬埔寨正在重新分类城市和农村地区。虽然这些重新分类已经起草，但尚未获得柬埔寨王国政府的批准。一旦获得批准和采用，新的分类将在未来的普查/调查中使用。
数据收集方式
---------------------------
面对面 [f2f]
研究工具
---------------------------
CIPS 2004的草案问卷大致遵循1998年总普查的模式。然而，进行了一些修改，增加了一些新问题，例如
(i) 0-14岁与母亲同住的孩子
(ii) 一个人是否的母亲还活着
(iii) 过去一年内家庭中死亡细节，重点关注孕产妇死亡率。

(i) 和 (ii) 中提到的问题分别旨在通过应用自己的孩子方法来估计生育率（通过孤儿方法来估计死亡率）。要包括的问题由柬埔寨2004年间普查人口调查工作组仔细考虑，该工作组的成员大多来自部委、非政府组织和国际机构。问卷于2003年11月由国家统计局工作人员在城乡（城市和农村）进行了两次现场测试。预测试的目的是对整个过程进行全面排练，特别是测试现场的问题，以便对措辞或定义进行纠正，并估计人口普查区域测绘、房屋登记、抽样和选定家庭普查所需的时间。根据预测试的经验，问卷被修改并最终确定。

CIPS 2004使用了两种类型的问卷：A表房屋清单和B表家庭问卷。

A表用于在调查之夜（2004年3月3日）之前的初步调查阶段收集有关包含一个或多个家庭的建筑的信息。收集的信息包括：墙壁、屋顶和地板的建筑材料、是否是全部或部分住宅建筑、建筑内的家庭数量、户主姓名和性别以及通常居住在家庭中的人数。

B表（分为五部分）用于参考时间之后的调查期间进行普查。

第一部分收集了调查之夜所选家庭通常成员、在场访客以及调查之夜缺席的通常成员的信息。

第二部分用于收集家庭每个通常成员和每个在场访客的信息。收集的信息包括：全名、与户主的关系、性别、年龄、亲生母亲、0-14岁与母亲同住的孩子、婚姻状况、首次结婚年龄、母语、宗教、出生地、以前居住地、停留时间、迁移原因、识字率、全日制教育和经济特征。

第三部分用于收集生育年龄的女性（15-49岁）以及这些女性所生的孩子的信息。

第四部分收集有关家庭状况和设施的信息：主要光源、主要烹饪燃料、是否提供厕所设施、主要饮用水来源以及家庭占用的客厅数量。

第五部分用于记录过去一年内家庭中死亡的信息：死者姓名、性别、与户主的关系、死亡时的年龄、死亡是否已向民事当局登记、死亡原因和孕产妇死亡率信息。

数据清理操作
---------------------------
完成的记录（A表、B表、I表、II表、地图和其他表格）由国家统计局调查协调员在规定日期从各省系统地收集，并提交给国家统计局的团队接待员。国家统计局调查协调员在3月7日至10日进行了培训，分成三个由两人组成的团队，以接收和整理已完成的形式和地图，在经过现场检查后进行处理。DUC规定了控制表格，以记录每个表格，没有任何遗漏。这些记录被仔细检查、注册并储存在记录室中。问卷的编辑和编码是手工完成的，之后将问卷提交给计算机部门进行进一步处理。编辑和编码的说明被修订和扩展。对高级工作人员进行了编辑和编码培训，他们反过来又必须培训其他编辑和编码人员。

编辑过程的目的在于消除明显不一致、不正确和不完整的事项，并提高收集到的数据质量。在出生地和以前居住地方面，必须非常仔细地进行编码，使用区和省代码，以及使用联合国国际标准职业分类（ISCO）和国际标准工业分类（ISIC）分别进行职业和行业的编码。为此，国家统计局利用了具有调查及其概念的良好知识和经验的员工。那些曾担任培训师或监督员的人被分配这项工作，并辅以经过良好培训和测试的员工。编辑和编码由两个团队（每个团队有六名编辑和一名团队领导）完成；因此，专门接受职业/行业编码培训的一名编辑将对第2部分家庭问卷的第20和22列进行编码。团队成员的工作完全由团队领导检查。编辑和编码培训于3月23日至26日进行。手动处理始于3月29日，并于2004年5月底完成。
响应率
---------------------------
响应率为95%。
抽样误差估计
---------------------------
对B表中的变量的一些总计、均值和比例的估计进行了抽样误差的计算。
用于计算的计算软件是STATA 8.0。对于此处呈现的计算，我们假设分层是在省份和城市/农村（对样本选择使用隐式57层省份和城市/农村分层）。

在45个层中的7个层中只有一个PSU（EA）被选中。这导致标准误差计算出现问题，在这些层中无法获得标准误差。在这些层中，我们将唯一的EA分为两部分，并将这两部分定义为两个PSU。

对于主要领域（如城市/农村和男性/女性）的估计，标准误差通常很小。在许多情况下，变异系数（CV）低于1%。对于省级估计，特别是对于样本量小的省份（例如省份19），变异系数显著较高。设计效应（Deff）已对一些估计进行了计算。正如预期的那样，对于人口特征估计，它们相当低。对于社会经济特征（如就业状况）的估计，设计效应相当高。对于“首次结婚年龄”和“婚姻状况”这样的人口特征，我们发现主要领域（如男性/女性和城市/农村）的设计效应低于5。社会经济特征通常比人口特征更“聚集”，这体现在通常较高的设计效应中。对于主要领域估计，我们发现设计效应高达20，偶尔有200或更高的非常高的值。这些“异常”值发生在样本量（PSU数量）小，以及PSU平均值（或比例）显示出较大变异的情况下。一个例子是估计城市地区政府雇员比例的设计效应为285。在这个领域，比例在102个PSU之间变动很大，范围从0%到75%。