National Panel Survey 2008-2009 - Tanzania

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

The NPS is nationally-representative household survey which provides measures of poverty, agricultural yields, and other key development indicators. The NPS is an “integrated” household survey, in that it covers a broad range of topics in the same questionnaire – from education and health to crime, gender-based violence and a range of other sections – to allow analysis of the links between sectors and the determinants of development outcomes.

The National Panel Survey (NPS) was designed to meet three principle objectives.

The first, overarching goals was to monitor progress toward the goals set out in the National Strategy for Growth and Poverty Reduction (aka, the MKUKUTA goals) and other national development objectives (MDG, PAF, etc.). The NPS provides high-quality, annual data on a long list of MKUKUTA indicators that is both nationally representative and comparable over time. As such, the NPS is intended to provide a key benchmark for tracking progress on poverty reduction and a wide range of other development indicators.

The second goal of the NPS is to facilitate better understanding of the determinants of poverty reduction in Tanzania. The NPS will enable detailed study of poverty dynamics at two levels. In addition to tracking the evolution of aggregate poverty numbers at the national level in years between Household Budget Surveys, the NPS will enable analysis of the micro-level determinants of poverty reduction at the household level. Panel data will provide the basis for analyzing the causal determinants of income growth, increasing or decreasing yields, improvements in educational achievement, and changes in the quality of public service provision over time by linking changes in these outcomes to household and community characteristics.

A third objective of the NPS is to provide data to evaluate the impact of specific policies and programs. With its national coverage and long time frame, the NPS will provide an ideal platform to conduct rigorous impact evaluations of government and non-government development initiatives. To achieve this goal, the National Bureau of Statistics will need to work in close collaboration with the relevant line ministries to link administrative data on relevant projects to changes in development outcomes measured in the survey.

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

The survey covered all regions and all districts of Tanzania, both mainland and Zanzibar.

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

Sample survey data [ssd]

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

In order to monitor progress toward the MKUKUTA goals, it was vital that the NPS have a nationally-representative sample design. As such, in 2008/09 the NPS interviewed 3,280 households spanning all regions and all districts of Tanzania, both mainland and Zanzibar.

The sample size of 3,280 households was calculated to be sufficient to produce national estimates of poverty, agricultural production and other key indicators. It will also be possible in the final analysis to produce disaggregated poverty rates for 4 different strata: Dar es Salaam, other urban areas on mainland Tanzania, rural mainland Tanzania, and Zanzibar. Alternatively, estimates of most key indicators can be produced at the zone level, as used for the Demographic and Health Survey (DHS) reports and other surveys. There are 7 of these zones in total on the mainland: North, Central, Eastern, South, Southern Highlands, West and Lake. As with any survey though, the confidence of the estimates declines as statistics are disaggregated into smaller zones.

Due to the limits of the sample size it is not possible to produce reliable statistics at the regional or district level.

The guiding principle in the choice of sample size, following standard practice for NBS surveys, was to produce estimates with a 95% confidence interval no larger than 5% of the mean for key indicators. In this case, household consumption and maize yields were used as the basis for those calculations.

The NPS was based on a stratified, multi-stage cluster sample design. The principle strata were Mainland versus Zanzibar, and within these, rural versus urban areas, with a special stratum set aside for Dar es Salaam. Within each stratum, clusters were chosen at random, with the probability of selection proportional to their population size. In urban areas a 'cluster' was defined as a census enumeration area (from the 2002 Population and Housing Census), while in rural areas an entire village was taken as a cluster. This primary motivation for using an entire village in rural areas was for consistency with the HBS 2007 sample which did likewise.

Based on the 2002 Population and Housing Census, rural residents comprise roughly 77% of the population, compared with 63% of the NPS sample. The NPS sample gives slighter greater weight to urban areas due to the higher levels of inequality in these areas, and added difficulty in estimating poverty rates and other statistics. Similarly, Zanzibar comprised roughly 3% of the Tanzanian population in the 2002 census, but constitutes nearly 15% of the NPS sample, so as to allow separate Zanzibar-specific estimates to be presented for most indicators.

Finally, although it has been stressed that the 2008/09 round is the first year of the NPS, the sample design for year 1 was deliberately linked to the 2007 HBS to facilitate comparison between the surveys. On mainland Tanzania, 200 of the 350 in the NPS were drawn from the 2007 HBS sample (this included all 140 rural HBS clusters). Within these 200 HBS clusters, a portion of the (8) households sampled for the NPS were taken from the sample of (24) HBS households in the cluster. (The number of HBS households sampled varied from cluster to cluster, in proportion to the share of the population, as measured through a comprehensive household listing, that had remained stationary in the cluster since the time of the HBS. This was done to ensure that the NPS sample remained nationally representative despite possible non-random attrition of HBS households.)

This design created a panel of approximately 1,200 HBS households - interviewed in both the HBS and NPS - within the total sample of 3,280 NPS households.

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

Face-to-face [f2f]

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

The main survey instrument of the NPS was the household questionnaire. This was administered to all households in the sample.

General household information – including food consumption and other household expenditure, which is central to poverty measurement – was solicited from the household head or another knowledgeable adult member of the household. In addition, wherever possible, each individual member over 5 years of age was interviewed directly for sections on education, health, labour, and food eaten outside the home.

In addition to the household questionnaire, a separate 46-page agricultural questionnaire was administered to all households with any agricultural activities (including farming, fishing or livestock, or ownership of any shamba even if not under cultivation). The agricultural questionnaire included detailed sections on each plot and each crop under cultivation, as well as information on farm assets, extension services, use and marketing of farm by-products, etc. For a sample of roughly 25% of the farming households, enumerators used GPS devices to directly measure the size of all farming plots.

Finally, apart from the questionnaires administered to households, a separate community questionnaire collected information from village, kitongoji and/or mtaa leaders. The community questionnaire covered topics including local administration and governance and access to basic services.

In a number of places, the NPS questionnaires provide extra detail relevant to MKUKUTA progress that goes beyond the specific indicators outlined in the MKUKUTA monitoring framework. In such cases, additional tables and statistics have been presented – in the relevant sections of the report – as a way of providing a deeper understanding of the process at work underlying progress on the core indicators. Key examples here are the enormous detail available on smallholder farming activities, which go far beyond the basic MKUKUTA indicators on technology usage and food production, and the in depth questions in the NPS on genderbased violence.

摘要
---------------------------

国民贫困状况调查（NPS）是一项具有全国代表性的家庭调查，旨在提供贫困、农业产量以及其他关键发展指标的衡量标准。NPS是一项‘综合’的家庭调查，因为它在同一份问卷中涵盖了广泛的主题——从教育、健康到犯罪、基于性别的暴力以及其他多个领域——以便分析各个部门之间的联系以及发展成果的决定因素。

国民面板调查（NPS）旨在实现三个主要目标。

第一个，总体目标是监测向《国家增长与减贫战略》（亦称MKUKUTA目标）以及其他国家发展目标（如MDG、PAF等）设定的目标进展情况。NPS提供了关于MKUKUTA一系列指标的高质量年度数据，这些数据既具有全国代表性，又能在时间上相互比较。因此，NPS旨在成为跟踪贫困减少以及其他一系列发展指标进展的关键基准。

NPS的第二个目标是促进对坦桑尼亚贫困减少决定因素的更好理解。NPS将使对贫困动态的细致研究成为可能，在两个层面上进行研究。除了在家庭预算调查之间的年份追踪全国层面的总体贫困演变外，NPS还将使分析家庭层面的贫困减少的微观决定因素成为可能。面板数据将为分析收入增长、产量增加或减少、教育成就的提高以及公共服务质量随时间的变化的因果决定因素提供基础，这些决定因素的变化与家庭和社区特征相联系。

NPS的第三个目标是提供数据以评估特定政策和计划的影响。凭借其全国覆盖范围和长期时间框架，NPS将为开展政府和非政府发展倡议的严格影响评估提供一个理想的平台。为了实现这一目标，国家统计局需要与相关主管部门紧密合作，将相关项目的行政数据与调查中测量的发展成果变化联系起来。

地理覆盖范围
---------------------------

调查涵盖了坦桑尼亚的所有地区和所有区，包括大陆和桑给巴尔。

数据类型
---------------------------

样本调查数据 [ssd]

抽样程序
---------------------------

为了监测向MKUKUTA目标的进展，NPS必须具有全国代表性的样本设计至关重要。因此，在2008/09年，NPS对坦桑尼亚所有地区和所有区——包括大陆和桑给巴尔——的3,280个家庭进行了访谈。

样本量为3,280个家庭，计算结果表明足以产生贫困、农业产量和其他关键指标的国家估计值。在最终分析中，还可能产生4个不同层级的人口贫困率：达累斯萨拉姆、坦桑尼亚大陆的其他城市地区、坦桑尼亚大陆的农村地区和桑给巴尔。此外，大多数关键指标的估计值可以在地区水平上产生，如人口统计与健康调查（DHS）报告和其他调查中所使用的那样。大陆上一共有7个这样的地区：北部、中部、东部、南部、南部高地、西部和湖地区。但是，与任何调查一样，随着统计数据被分解到更小的地区，估计值的置信度会下降。

由于样本量的限制，不可能在地区或区一级产生可靠的统计数据。

选择样本量的指导原则，遵循国家统计局调查的标准做法，是产生估计值，其95%的置信区间不超过关键指标均值的5%。在这种情况下，家庭消费和玉米产量被用作计算的基础。

NPS基于分层、多阶段集群样本设计。主要层级是大陆与桑给巴尔，在这些层级中，农村与城市地区，特别为达累斯萨拉姆预留了一个特殊层级。在每个层级内部，集群是随机选择的，选择概率与它们的 人口规模成比例。在城市地区，‘集群’被定义为2002年人口和住房普查中的普查登记区域，而在农村地区，整个村庄被视为一个集群。在农村地区使用整个村庄作为集群的初级动机是为了与2007年HBS样本的一致性。

根据2002年人口和住房普查，农村居民约占人口的77%，而NPS样本中约占63%。NPS样本略微增加了对城市地区的权重，因为这些地区的差异程度较高，估计贫困率和其他统计数据更加困难。同样，桑给巴尔在2002年人口普查中约占坦桑尼亚人口的3%，但在NPS样本中占近15%，以便为大多数指标提供单独的桑给巴尔特定估计值。

最后，尽管已经强调2008/09轮是NPS的第一年，但第1年的样本设计故意与2007年HBS相联系，以促进调查之间的比较。在坦桑尼亚大陆，NPS中的350个样本中有200个来自2007年HBS样本（这包括所有140个农村HBS集群）。在这些200个HBS集群中，部分NPS样本的（8）个家庭是从集群中的（24）个HBS家庭样本中抽取的。（HBS家庭的样本数量根据集群而变化，与通过全面的家庭列表测量的自HBS以来在集群中保持静止的人口份额成比例。这是为了确保NPS样本在可能的HBS家庭非随机流失的情况下仍然具有全国代表性。

这种设计在3,280个NPS家庭样本中创造了一个大约1,200个HBS家庭的面板——这些家庭在HBS和NPS中都被采访过。

数据收集方式
---------------------------

面对面 [f2f]

研究工具
---------------------------

NPS的主要调查工具是家庭问卷。它被分发给样本中的所有家庭。

一般家庭信息——包括食物消费和其他家庭支出，这是贫困测量的核心——是从家庭户主或家庭中另一位了解情况的成年人那里获得的。此外， wherever possible，5岁以上的每个成员都直接接受教育、健康、劳动和在家外食用的食物部分的访谈。

除了家庭问卷外，还对从事任何农业活动（包括农业、渔业或牲畜，或拥有任何shamba即使没有耕种）的家庭进行了单独的46页农业问卷。农业问卷包括关于每个地块和每个种植作物的详细部分，以及关于农场资产、推广服务、农场副产品的使用和营销等信息。对于大约25%的农业家庭样本，调查员使用GPS设备直接测量所有农业地块的面积。

最后，除了对家庭进行的问卷外，还进行了单独的社区问卷，收集来自村庄、kitongoji和/或mtaa领导者的信息。社区问卷涵盖了包括地方行政和治理以及基本服务获取在内的主题。

在许多地方，NPS问卷提供了关于MKUKUTA进展的额外细节，这些细节超越了MKUKUTA监测框架中概述的具体指标。在这种情况下，已在报告的相关部分中提供了额外的表格和统计数据，以便更深入地了解在核心指标进展过程中起作用的机制。关键示例包括关于小农户农业活动的详细信息，这些信息远远超出了MKUKUTA关于技术使用和食品生产的基本指标，以及NPS中对基于性别的暴力的深入问题。
}