Living Conditions Monitoring Survey VI 2010 - Zambia

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

The main objective of the 2006 and 2010 LCMS surveys was to provide the basis for comparison of poverty estimates derived from cross-sectional survey data between 2006 and 2010.

In addition, the survey provides a basis on which to:
- Monitor the impact of government policies on the well being of the Zambian population.
- Monitor the level of poverty and its distribution in Zambia.
- Provide various users with a set of reliable indicators against which to monitor
- Identify vulnerable groups in society and enhance targeting in policy implementation.

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

In the LCMS 2010, all the 1000 sampled SEAs were enumerated representing 100 percent coverage at national level.

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

- Households
- Individuals

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

The survey covered all de jure household members (usual residents) resident in the household.

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

Sample survey data [ssd]

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

Sample stratification and allocation
The sampling frame used for the LCMS VI was developed from the 2000 Census of Population and Housing. The country is administratively demarcated into 9 provinces, which are further divided into 72 districts. The districts are further subdivided into 150 constituencies, which are in turn divided into wards. For the purposes of conducting CSO surveys, Wards are further divided into Census Supervisory Areas (CSA), which are further subdivided into Standard Enumeration areas (SEAs). For the purposes of this survey, SEAs constituted the Primary Sampling Units (PSUs). In order to have reasonable estimates at district level and at the same time take into account variation in the sizes of the districts, the survey adopted the Square Root sample allocation method, (Leslie Kish, 1987). This approach offers a compromise between equal and proportional allocation i.e. small sized strata (Districts) are allocated larger samples compared to proportional allocation. However, it should be pointed out that the sample size for the smallest districts is still fairly small, so it is important to examine the confidence intervals for the district-level estimates in order to determine whether the level of precision is adequate. The allocation of the sample points to rural and urban strata was done in such a way that it was proportional to their sizes in each district. Although this method was used, it was observed from the LCMS 2006 that the coefficient of variation (CV) of the poverty estimates was highest in districts which are predominantly urban and lowest in rural districts. This means that the sample size in some urban districts may have been inadequate to measure poverty with a good level of precision. That is, given the higher variability in the urban districts, a larger sample size would be required. Also some districts had very low CV estimates, indicating a higher level of precision for the poverty estimates. In order to try and improve the precision of the poverty estimates for the urban districts, the initial distribution of the sample was adjusted. It was necessary to increase the number of PSUs for some districts without increasing the budget and at the same time not compromising significantly the precision of the poverty estimates for rural areas. Rural districts which had the lowest CVs in the 2006 LCMS results had their sample size reduced, and these were in turn distributed to districts with the highest CVs. The distribution of the sample for the LCMS 2006 and LCMS 2010 were initially the same but changed after the later was adjusted. Table 2.1 in the Survey Report shows the allocation of PSUs in the survey.

Sample Selection
The LCMS VI employed a two-stage stratified cluster sample design whereby during the first stage, 1000 SEAs were selected with Probability Proportional to Estimated Size (PPES) within the respective strata. The size measure was taken from the frame developed from the 2000 Census of Population and Housing. During the second stage, households were systematically selected from an enumeration area listing. The survey was designed to provide reliable estimates at the district, provincial, rural/urban and national levels. However, the reliability for some indicators may be limited for the smaller districts, given the limited sample size. This will be determined by the tabulation of sampling errors and confidence intervals.

Selection of households
Listing of all the households in the selected SEAs was done before a sample of households to be interviewed was drawn. In the case of rural SEAs, households were stratified and listed according to their agricultural activity status. Therefore, there were four explicit strata created at the second sampling stage in each rural SEA namely, the Small Scale Stratum (SSS), the Medium Scale Stratum (MSS), the Large Scale Stratum (LSS) and the Non-agricultural Stratum (NAS). For the purposes of the LCMS VI, Seven, five and three households were selected from the SSS, MSS and NAS, respectively. The large scale households were selected on a 100 percent basis. The urban SEAs were explicitly stratified into low cost, medium cost and high cost areas according to CSO's and local authority classification of residential areas. From each rural and urban SEA, 15 and 25 households were selected, respectively. However, the number of rural households selected in some cases exceeded the prescribed sample size of 15 households depending on the availability of large scale farming households.The selection of households from various strata was preceded by assigning fully responding households sampling serial numbers. The circular systematic sampling method was used to select households. The method assumes that households are arranged in a circle (G. Kalton, 1983) and the following relationship applies:
Let N = nk,
Where:
N = Total number of households assigned sampling serial numbers in a stratum
n = Total desired sample size to be drawn from a stratum in an SEA
k = The sampling interval in a given SEA calculated as k=N/n.

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

Face-to-face [f2f]

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

Three types of questionnaires will be used in the survey. These are:-
1. The Listing Booklet - to be used for listing all the households residing in the selected Standard Enumeration Areas (SEAs)
2. The Main questionnaire - to be used for collecting detailed information on all household members in the selected households
3. The Prices questionnaire:- to be used to collect unit prices of various commodities. This information is vital for harmonising regional differences in prices

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

The Living Conditions Monitoring Survey data were entered using CSPro version 4.0 software. The LCMS 2010 application used a double entry system unlike the LCMS 2006 application which used single entry. The 2010 data entry was done by two teams, one team in the Provinces and another one at CSO headquarters. The data were then compared and matched by a team of matchers. Errors identified by matchers were corrected as a way of completing data entry. The major advantage of double entry (verification) is that data entry errors generated by the data entry operator are greatly minimized. The data were then exported to SAS, SPSS and Stata formats for data cleaning bulation and analysis.

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

The household response rate was calculated as the ratio of originally selected households with completed interviews over the total number of households selected. The household response rate was also generally very high with a national average of 98 percent of the originally selected households for both survey periods.

摘要
---------------------------
2006年和2010年LCMS调查的主要目标是提供比较2006年至2010年间基于横断面调查数据得出的贫困估计的基础。
此外，调查还提供了以下基础：
- 监测政府政策对赞比亚人民福祉的影响。
- 监测赞比亚的贫困水平和分布。
- 为各种用户提供一套可靠的指标，以供监测之用。
- 识别社会中的脆弱群体，并提高政策实施中的目标定位。
地理覆盖范围
---------------------------
在LCMS 2010中，对所有1000个样本监督区域进行了统计，代表国家层面的100%覆盖率。
分析单元
---------------------------
- 家庭
- 个人
总体
---------------------------
调查涵盖了所有法定家庭成员（常住居民）。
数据类型
---------------------------
样本调查数据 [ssd]
抽样程序
---------------------------
样本分层和分配
用于LCMS VI的抽样框架是从2000年人口和住房普查中开发的。国家被行政上划分为9个省，这些省进一步划分为72个区。区进一步划分为150个选区，选区再进一步划分为街道。为了进行CSO调查，街道被进一步划分为普查监督区域（CSA），CSA再进一步划分为标准统计区域（SEAs）。为了本调查的目的，SEAs构成了主要抽样单位（PSU）。为了在区级获得合理的估计值，同时考虑到各区大小的差异，调查采用了平方根样本分配方法（Leslie Kish，1987年）。这种方法在平等分配和比例分配之间提供了一个折中方案，即小规模层（区）分配的样本量比比例分配大。然而，应该指出的是，最小区的样本量仍然相当小，因此检查区级估计的置信区间，以确定精度是否足够是很重要的。样本点到农村和城市层的分配是以它们在每个区的大小成比例进行的。尽管使用了这种方法，但LCMS 2006观察到贫困估计值的变异系数（CV）在以城市为主的区最高，在乡村区最低。这意味着一些城市区的样本量可能不足以以良好的精度衡量贫困。也就是说，鉴于城市区的更高变异性，需要更大的样本量。此外，一些区的CV估计值非常低，表明贫困估计值的精度更高。为了尝试提高城市区的贫困估计值的精度，对样本的初始分配进行了调整。有必要在不增加预算的同时，增加某些区的PSU数量，同时不显著降低农村地区的贫困估计值的精度。在2006年LCMS结果中CV最低的农村区减少了样本量，这些样本随后分配到CV最高的区。LCMS 2006和LCMS 2010的样本分配最初是相同的，但在后期进行了调整。调查报告中的表2.1显示了调查中PSU的分配。
样本选择
LCMS VI采用两阶段分层聚类样本设计，在第一阶段，在各自的层中，以概率比例于估计规模（PPES）选择了1000个SEAs。规模衡量标准来自2000年人口和住房普查中开发的框架。在第二阶段，从名录中系统地选择了家庭。该调查旨在提供可靠的区级、省级、农村/城市和国家级别的估计值。然而，鉴于样本量有限，某些较小区的某些指标的可靠性可能受到限制。这将通过抽样误差和置信区间的编制来确定。
家庭选择
在从选定的SEAs中抽取要访谈的家庭样本之前，对所有选定的SEAs中的家庭进行了列表。在农村SEAs的情况下，家庭根据其农业活动状况进行了分层和列表。因此，在第二个抽样阶段，每个农村SEAs中创建了四个明确的层，即小规模层（SSS）、中规模层（MSS）、大规模层（LSS）和非农业层（NAS）。为了LCMS VI的目的，从SSS、MSS和NAS中分别选择了七个、五个和三个家庭。大规模家庭以100%的比例选择。城市SEAs明确地分层为低成本、中成本和高成本地区，根据CSO和地方当局对住宅区的分类。从每个农村和城市SEAs中分别选择了15和25个家庭。然而，在某些情况下，农村家庭的选择数量超过了规定的15个家庭的样本量，这取决于大规模农业家庭的可获得性。在从各种层中选择家庭之前，为完全响应家庭分配了采样序列号。使用了圆形系统抽样方法来选择家庭。该方法假设家庭是按环形排列的（G. Kalton，1983年），以下关系适用：