Welfare Monitoring Survey 2000 - Ethiopia

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

Ethiopia is one of those countries that suffer the hardest hits of poverty. Persistent war and drought and inappropriate policies are presumed to enhance the extent of poverty in the country. According to the report on Poverty Situation in Ethiopia which was based on the 1995/96 Household Income, Consumption and Expenditure Survey and the 1996 Welfare Monitoring Survey 45.5 percent of the total population are found to live below the poverty line. The report has also revealed the disparity among urban-rural residents in which 47.5 percent of the urban population. Hence, the issue of poverty reduction would necessarily be an agenda of higher priority for the government and policy markers.

As in the case of a number of Africa countries that undertook the Social Dimensions of Adjustment (SDA) program, the issue of welfare monitoring in Ethiopia arose as part of the Economic Reform Program (ERP) currently being undertaken in the country. The ERP specifically and strongly underlies the effects of the reform program on poverty and the analytical capacity of the government to monitor such effects. To this end, the government has set up a Welfare Monitoring System (WMS) by mid 1994.

In view of the wider context of the problem, the establishment of the WMS is envisaged to consist of the following major elements:
- establish an information system that provides a continuous picture of the poverty scenario in the country;
- indicate the impact of reform programs on the level of household welfare;
- establish follow-up procedures on the various programs and activates targeted towards poverty alleviation; and
- conduct regular statistical survey to assess, in particular, the efficiency of targeted programs.

In order to fulfill the data needs to monitor households' socioeconomic welfare and the ongoing economic reforms, the Central Statistical Authority (CSA) has been conducting Welfare Monitoring Surveys starting from 1996. Reports of the 1996 and 1998 Welfare Monitoring Surveys have also been disseminated.

The Welfare Monitoring Survey (WMS) 2000, like the previous ones, focuses on wide range of socioeconomic indicators, which are vital inputs in the process of monitoring and evaluation of policies, particularly in poverty reduction strategies. The report is presented in two volumes. Volume I presents results based on individual data base and Volume II presents the findings based on household database. Proxy estimate of households' domestic expenditure obtained by recall interview (with reference periods of 7 days and a month prior to the data of interview) is used to classify households (on quintile basis) for the purpose of tabulating the results.

Objectives of the Welfare Monitoring System
The WMS which involves various ministries and the Central Statistical Authority (CSA) is established with the following objectives:
- provide baseline data on existing poverty situation and establish a system of information gathering on relevant key indicators;
- identify poor and vulnerable group that could be the focus of targeted intervention programs;
- undertake periodic surveys and researches to evaluate targeted programs;
- assess the short and medium term effects of macroeconomic and sectorial policies and programs on the poor;
- produce conclusive reports and suggestions needed for due attention by the government and concerned implementing agencies.

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

The WMS 2000 covered the population in sedentary areas of the country on a sample basis excluding the non-sedentary population in Afar and Somalia Regional States. That is, the survey covered the population in sedentary areas of the nine Regional States and two administrative regions, each of which is composed of rural and urban parts.

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

- Households
- Individual (including adult women aged 15 and above, children aged 5 and below)

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

The survey covered households in the selected samples except residents of collective quarters, homeless persons and foreigners.

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

Sample survey data [ssd]

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

The WMS 2000 covered both the urban and the sedentary rural parts of the country. The survey has not covered six zones in Somalia Regional State and two zones in Afar Regional State that are inhabited mainly by nomadic population. For the purpose of the survey, the country was divided into three categories. That is, the rural parts of the country and the urban areas that were divided into two broad categories taking into account sizes of their population.

Category I: Including rural area of 44 zones in 7 regions, 5 special weredas in SNNPR and rural areas of Gambella, Harari, Addis Ababa and Dire Dawa regions each of which are survey domains (reporting levels).The regions that constitute the 44 zones are Tigray, Afar, Amhara, Oromiya, Somalia, Benishangul_Gumuz, and SNNPR. All in all 54 basic rural domains including total rural (country level) are defined for the survey.

Category II: Comprises if all regional capitals and five other urban centers. Each urban center in this category is the survey domain (reporting level) for which separate survey results for major survey characteristics are reported.

Category III: Urban centers in the country other than those under category II are grouped to this category. There are four domains (reporting levels) in this category: Tigray other urabn, Amhara other urban, Oromiya other urban and SNNPR other urban. Eleven additional domains other than those reporting levels defined in Category II and Category III, can be constructed by combining basic domains from these two categories. These domains are:
1) Tigray urban, 2) Afar urban, 3) Amhara urban, 4) Oromiya urban, 5) Somalia urban, 6) Beneshangul-Gumuz urban, 7) SNNPR urban, 8) Gambella urban, 9) Harari urban, 10) Addis Ababa urban and 11) Dire Dawa urban

In addition to the above urban and rural domains, survey results can also be reported at regional and country levels by aggregating the corresponding survey results for urban and rural areas.

Definition of the survey domains was based on both technical and resource considerations. More specifically, sample size for the domains were determined to enable provision of major indicators with reasonable precision subject to the resources that were available for the survey.

Selection Scheme and Sample Size in Each Category
a) Category I: A stratified two-stage sample design was used to select the sample in which the primary sampling units (PSUs) were EAs. Sample enumeration area (EAs) from each domain were selected using systematic sampling that is probability proportional to size, size being number of households obtained from 1994 population and housing census. A total of 1450 EAs were selected form the rural parts of the country. Within each sample EA a fresh list of households was prepared at the beginning of the survey's fieldwork and for the administration of the survey questionnaire 12 households per sample EA for rural areas were systematically selected.

b) Category II: In this category also, a stratified two-stage sample design was used to select the sample. In this category a strata constitutes all the Regional State Capitals and the five Major Urban Centers in the country. The primary sampling units (PSUs) are the EA's in the Regional State Capitals and the five Major Urban Centers and excludes the special EAs (non-conventional households). Sample enumeration areas (EAs) from each strata were selected using systematical sampling that is probability proportional to size, size being number of households obtained from the 1994 population and housing census. A total of 373 EAs were selected from this domain of study. Within each sample EAs a fresh list of households was prepared at the beginning of the survey's field work and for the administration of the questionnaire 16 households per sample EA were systematically selected.

c) Category III: Three-stage stratified sample design was adopted to select the sample from domains in category III. The PSUs were other urban centers selected using systematic sampling that is probability proportional to size; size being number if households obtained from the 1994 population and housing census. The secondary sampling units (SSUs) were EAs which were selected using systematic sampling that is probability proportional to size; size being number of households obtained from the 1994 population and housing census. A total of 169 EA's selected from the sample of other urban centers and was determined by proportional allocation to their size of households from the 1994 census. Ultimately, 16 households within each of the sample EAs were selected systematically from a fresh list of households prepared at the beginning of the survey's fieldwork for the administration of the survey questionnaire.

Note: Distribution of EAs and households covered in the survey by domain (reporting levels) and category are given in Table II.1 and Table II.2 of 2000 Welfare Monitoring Survey report which is provided in this documentation.

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

Face-to-face [f2f]

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

Basically there were two types of questionnaires; one referring to individual household members and the other pertaining to households in general.
1. Individual level questionnaires were used to collect basic population characteristics, health, education, on nutritional status of the children (anthropometric measurements) and immunization.
2. Household-based questionnaires included modules on housing amenities, accessibility of basic facilities such as food market, post office and telephone, possession of household asset and schedule on living standard indicators with respect to basic necessities (food, clothing and shelter). Proxy measure of household expenditure was also included in WMS questionnaires to hint at the economic situation of households.

Here are list of forms of the questionnaires:
Form 1: Economic and Demographic Characteristics
Form 2: Educational Status
Form 3: Heath
Form 4: Anthropometry, Immunization and Child Care
Form 5A-B: Housing Standard and Amenities
Form 6: Access to Facilities
Form 7: Household Assets
Form 8A: Household Expenditure
Form 8B: Household Income
Form 9: Indicators on Living Standard

Note: The questionnaires are provided as external resource.

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

After the completion of the fieldwork the filled-in questionnaires were retrieved from the branch statistical offices for data processing. The first stage data processing activity was training of data editors and coders which was held at the head office by subject matter department staff. About 55 editors-coders and 8 verifiers took part in the manual editing, coding and verification activities, which lasted for about a month. Data entry took about 15 days using 28 computers and as many data encoders.

Machine data cleaning, data weighting with proper sampling weights and tabulation activities were carried out procedurally by the professional staff from involved departments at the head office. The Integrated Microcomputer Processing System (IMPS) software was used for data entry, consistency checks and tabulation of survey results.

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

A total of 1,992 Enumeration Areas (1450 in rural and 542 in urban) were selected to be covered in the survey in all regions. The survey succeeded to cover 1984 EAs (99.6 5) of the selected EAs in the rural and urban areas.

With regard to ultimate sampling units, it was planned to cover a total of 26,072 households (17,400 in rural and 8,672 in urban areas) all over the country. The response rate was about 100 percent (99.34 percent or 17,285 households in rural and 99.67 percent or 8,643 households in urban areas). Only 115 of the selected rural households and 29 of the selected urban households refused to cooperate to provide information for survey questionnaire.

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

Details of the estimation procedures, standard errors and coefficients of variations of selected variables are presented in Annex II and Annex III, respectively of Welfare Monitoring Survey 2000 report which is provided as external resources.

摘要
---------------------------
埃塞俄比亚是那些遭受贫困重创的国家之一。持续的战争、干旱及不当政策被认为是加剧该国贫困程度的重要因素。根据基于1995/96年家庭收入、消费和支出调查以及1996年福利监测调查的《埃塞俄比亚贫困状况报告》，发现有45.5%的总人口生活在贫困线以下。报告还揭示了城乡居民之间的差异，其中47.5%的城市人口。因此，贫困减少问题必然成为政府及政策制定者优先考虑的议程。

与许多实施社会调整维度（SDA）计划非洲国家的情况类似，埃塞俄比亚的福利监测问题作为当前国家正在进行的经济改革计划（ERP）的一部分而出现。ERP特别强调改革计划对贫困的影响以及政府分析这种影响的能力。为此，政府在1994年中叶建立了福利监测系统（WMS）。

鉴于问题的广泛背景，WMS的建立设想包含以下主要要素：
- 建立一个信息系统，提供国家贫困状况的持续图景；
- 指出改革计划对家庭福利水平的影响；
- 建立针对贫困缓解的各类项目和活动的后续程序；
- 定期进行统计调查，以评估尤其是针对项目的效率。

为了满足监测家庭社会经济福利和持续经济改革的数据需求，中央统计局（CSA）自1996年起开始进行福利监测调查。1996年和1998年福利监测调查的报告也已公布。

2000年的福利监测调查（WMS）与之前类似，重点关注广泛的社会经济指标，这些指标是监测和评估政策过程，尤其是在贫困减少策略中的关键输入。报告分为两卷。第一卷基于个人数据库呈现结果，第二卷基于家庭数据库呈现发现。通过回忆访谈（参考期限为数据收集前的7天和一个月）获得的家庭国内支出代理估计用于根据五分位数对家庭进行分类，以便编制结果。

福利监测系统目标

涉及多个部和中央统计局（CSA）的WMS旨在实现以下目标：
- 提供现有贫困状况的基线数据，并建立有关相关关键指标的信息收集体系；
- 识别贫困和脆弱群体，作为针对性干预计划的重点；
- 定期进行调查和研究，以评估针对性项目；
- 评估宏观经济和行业政策及计划对贫困人群的短期和中期影响；
- 生成政府及相关部门应予关注的结论性报告和建议。

地理覆盖范围
---------------------------
2000年的WMS覆盖了国家定居地区的居民，以抽样为基础，排除了阿法尔和索马里地区州的非定居人口。也就是说，调查覆盖了九个地区州和两个行政区域的定居地区，每个区域都由农村和城市部分组成。

分析单元
---------------------------
- 家庭
- 个人（包括15岁及以上的成年女性、5岁及以下的儿童）

总体
---------------------------
调查涵盖了所选样本中的家庭，但不包括集体宿舍居民、无家可归者和国外人士。

数据类型
---------------------------
样本调查数据 [ssd]

抽样程序
---------------------------
2000年的WMS覆盖了国家的城市和定居农村地区。调查未涵盖索马里地区州的六个区域和阿法尔地区州的两个区域，这些区域主要居住着游牧民族。为了进行调查，国家被分为三个类别。即国家的农村部分和被分为两大类别的城市地区，这些类别考虑到它们的人口规模。

类别I：包括7个地区的44个区域的农村地区、SNNPR的5个特殊地区和冈比亚拉、哈拉里、亚的斯亚贝巴和迪雷德瓦地区农村地区，每个地区都是调查领域（报告级别）。构成44个区域的地区是提格雷、阿法尔、阿姆哈拉、奥罗莫、索马里、本尚古勒-古穆兹和SNNPR。总共定义了54个基本农村领域，包括总农村（国家级别）领域，用于调查。

类别II：包括所有地区首府和五个其他城市中心。本类别中的每个城市中心都是调查领域（报告级别），其中单独报告了主要调查特征的调查结果。

类别III：除类别II中的城市以外的国家其他城市中心被归入本类别。本类别中有四个领域（报告级别）：提格雷其他城市、阿姆哈拉其他城市、奥罗莫其他城市和SNNPR其他城市。可以通过结合这两个类别的基本领域构建11个额外的领域，这些领域是：
1) 提格雷城市、2) 阿法尔城市、3) 阿姆哈拉城市、4) 奥罗莫城市、5) 索马里城市、6) 本尚古勒-古穆兹城市、7) SNNPR城市、8) 冈比亚拉城市、9) 哈拉里城市、10) 亚的斯亚贝巴城市和11) 迪雷德瓦城市。

除上述城市和农村领域外，还可以通过汇总城市和农村地区的相应调查结果，在地区和国家级别报告调查结果。

调查领域的定义基于技术和资源考虑。具体而言，领域样本量是根据能够以合理的精度提供主要指标而确定的，前提是调查可用的资源。

每个类别的选择方案和样本量
a) 类别I：使用分层两阶段样本设计选择样本，其中一级抽样单位（PSU）是选举区。使用系统抽样从每个领域选择选举区，即根据大小进行概率成比例抽样，大小是从1994年人口和住房普查中获得的家庭数量。总共从国家的农村部分选择了1450个选举区。在样本选举区内部，在调查的现场工作开始时和进行调查问卷管理时，为农村地区系统性地选择了每个样本选举区的12个家庭。

b) 类别II：在本类别中，也使用分层两阶段样本设计选择样本。在本类别中，一个层由所有地区首府和国家的五个主要城市中心组成。一级抽样单位（PSU）是地区首府和五个主要城市中心中的选举区，并排除了特殊选举区（非常规家庭）。使用系统抽样从每个层中选择选举区，即根据大小进行概率成比例抽样，大小是从1994年人口和住房普查中获得的家庭数量。总共从研究领域的本领域选择了373个选举区。在每个样本选举区内部，在调查的现场工作开始时和进行问卷管理时，为每个样本选举区系统性地选择了16个家庭。

c) 类别III：采用三阶段分层样本设计从类别III的领域选择样本。PSU是使用系统抽样选择的其它城市中心，即根据大小进行概率成比例抽样，大小是从1994年人口和住房普查中获得的家庭数量。二级抽样单位（SSU）是使用系统抽样选择的选举区，即根据大小进行概率成比例抽样，大小是从1994年人口和住房普查中获得的家庭数量。总共从其它城市中心的样本中选择了169个选举区，这是根据从1994年普查中获得的家庭数量按比例分配到其大小确定的。最终，从样本选举区的每个样本家庭中系统性地选择了16个家庭，这些家庭在调查的现场工作开始时和进行调查问卷管理时，为问卷管理而准备的新家庭清单。

注：有关调查覆盖的选举区和家庭以及按领域（报告级别）和类别划分的分布，请参阅本文件提供的2000年福利监测调查报告的第II.1表和第II.2表。

数据收集方式
---------------------------
面对面 [f2f]

研究工具
---------------------------
基本上有两种类型的问卷；一种涉及个人家庭成员，另一种涉及一般家庭。
1. 个人层面问卷用于收集基本人口特征、健康、教育、儿童营养状况（人体测量）和免疫。
2. 家庭层面问卷包括有关住房便利、基本设施（如食品市场、邮局和电话）的可及性、家庭资产拥有情况以及与基本需求（食物、衣物和住所）相关的生活标准指标模块。WMS问卷还包括家庭支出的代理测量，以暗示家庭的经济状况。

以下列出了问卷的形式：
表1：经济和人口统计特征
表2：教育状况
表3：健康
表4：人体测量、免疫和儿童护理
表5A-B：住房标准和便利设施
表6：设施可及性
表7：家庭资产
表8A：家庭支出
表8B：家庭收入
表9：生活标准指标

注：问卷作为外部资源提供。

数据清理操作
---------------------------
在完成现场工作后，填写好的问卷从分支机构统计办公室收回，以进行数据处理。第一阶段数据处理活动是数据编辑人员和编码人员的培训，这由总部的主管部门工作人员在总部举行。大约55名编辑人员和编码人员以及8名验证人员参加了为期约一个月的手动编辑、编码和验证活动。数据录入耗时约15天，使用28台计算机和同等数量的数据编码人员。

由总部涉及部门的职业工作人员按程序执行机器数据清理、数据加权（使用适当的抽样权重）和编制活动。集成微机处理系统（IMPS）软件用于数据录入、一致性检查和调查结果的编制。

响应率
---------------------------
总共选择了1,992个选举区（农村地区1450个，城市地区542个）以在所有地区进行调查。调查成功覆盖了所选的1984个选举区（农村地区99.6%），城市地区99.6%。

关于最终抽样单位，计划覆盖全国范围内的26,072个家庭（农村地区17,400个，城市地区8,672个）。响应率约为100%（农村地区99.34%，或17,285个家庭；城市地区99.67%，或8,643个家庭）。只有115个农村家庭和29个城市家庭拒绝合作，不为调查问卷提供信息。

抽样误差估计
---------------------------
关于估计程序、标准误差和选定变量的变异系数的详细信息分别见福利监测调查2000年报告的附件II和附件III，这些报告作为外部资源提供。