Ageing, Well-being and Development Project 2002-2008 - Brazil, South Africa

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

The purpose of the Ageing, Wellbeing and Development Project (Brazza2) was to investigate the impact on poverty and vulnerability within beneficiary households in Brazil and South Africa of grants, social pensions and the like. The survey aimed to help researchers interrogate the extent to which social assistance was enhancing quality of life, and whether income from old-age pensions and other social grants enhanced the material and perceived well-being of social pensioners and members of households.The study also inquired into perceptions of fortune and misfortune, to provide clues to the role of social assistance in boosting poorer households' resilience and their independence from the State.

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

Households and individuals

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

South Africa: the survey covered all members of black households in the rural Eastern Cape and black and coloured households in urban Western Cape.

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

Sample survey data

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

South Africa:
In South Africa, a company called Development Research Africa were commissioned to conduct the data collection. To conduct the sampling for this, they requested a list of EAs from Stats SA that satisfied the following criteria:

1. Predominantly black or coloured EAs
2. Predominantly defined (by Stats SA) as urban (formal or informal) in the Western Cape
3. Predominantly defined (by Statssa) as tribal or semi urban in the Eastern Cape; and
4. Did not contain institutions or farming areas (these EAs were excluded)

These CEAs were sent to DRA in several excel spreadsheets under the following headings for each magisterial district:

1. Geographical areas by population group of head of household for person weighted (African/Black or Coloured)
2. Geographical areas by enumeration area type for person weighted (rural: tribal villages, urban: formal or urban: informal)
3. Geographical areas by age for person weighted (56 years and older)
4. Geographical areas for household weighted (which provided the total number of households per CEA).

These data files were collated and then merged into three separate spreadsheets reflecting the respondent categories. All CEAs containing less than eighty households were deleted to further ensure that institutions or farming areas (as well as urban areas in the Eastern Cape) would not become eligible and also to limit the possibility of selecting CEAs with no eligible respondent households. These three databases became the three sample frames used to select the sample.

All the remaining CEAs were sorted in ascending order. A PSS sampling method was used to select the sample. This means that CEAs with a larger number of households have a greater chance of being selected into the sample. The two CEAs directly below the selected EAs were included as possible substitutions. Once the EA numbers were selected the maps were sourced from Stats SA. Only then could one determine the location of these CEAs. Because of the PPS methodology, EAs from smaller magisterial districts fell short of being selected into the sample whilst larger magisterial districts had more than one EA selected. In the Western Cape, the EAs could relatively easily be found on Cape Town street maps.

Twenty clusters or EAs were selected per respondent category. The target per category was about 333 interviews. It follows that about 17 interviews (333/20=17) had to be done per CEA. The desired number of households that need to be approached in a cluster or EA was the segment size. The segment size was dependent on the percentage of households that contain at least one person aged 55 years and over and on the response rate assumed. The segment size for each of the CEAs in the sample was calculated individually. For example, if 33 persons aged 55 or older resided in the CEA with 120 households and assuming a 95% response rate, 59 households would have to be approached (17/(15/120)*0.95) in the CEA in order to obtain 17 successful interviews per CEA. One limitation to the study here was that this formula does not take into consideration the possibility of two or more persons in this age category residing in a household.

Once the maps were acquired from Stats SA, they were verified and updated by the fieldworker through identifying the EA boundaries and by entering any features or changes to the map. The number of households were then counted and divided into segments with approximately equal number of households. One calculates the number of segments by dividing the segment size (described in the previous paragraph) by the actual number of households found and recorded in the EA. Some EAs may have only one segment (if segment size > total number of households in EA) or may have as many as five or six segments. One segment is then randomly selected. All the households in a particular segment were approached and all target households identified and surveyed. Finally, within the households, the person most knowledgeable about how money is spent in the household was selected as the first respondent. Thereafter all individuals 55 years of age and over were interviewed. The fieldworkers had to make three visits per household where the respondents were not available to maximize the possibility that the interview would be completed with the selected respondent. The project manager monitored the number of completed interviews. In instances where it seemed that the overall target of 333 interviews per respondent category area was unlikely, the fieldworkers had to survey the whole EA.

The twenty randomly-selected EAs in the rural Eastern Cape were located in the former Transkei and Ciskei 'homelands' in the magisterial districts of Zwelitsha, Keiskammahoek, Engcobo, Idutywa, Kentani, Libode, Lusikisiki, Mqanduli, Ngquleni, Nqamakwe, Port St Johns, Qumbu, Cofimvaba, Tabankulu, Tsomo, Willowvale and Lady Frere. The twenty randomly-selected EAs in the Cape Town metropole targeting urban black households were located in the magisterial districts of Goodwood, Wynberg, Mitchell's Plain (which includes the sprawling township of Khayelitsha) and Kuils River. The twenty randomly-selected EAs targeting urban coloured households were located in the same magisterial districts in Cape Town metropole as those targeting urban black households with the addition of Bellville.

The 2002 sample design prescribed that all households selected in the last stage, in the EA segment, had to be interviewed. As a result, a larger sample size was achieved in 2002 than the originally planned sample of 1000 interviews. A total of 1111 interviews was realised in 2002: 374 in rural black households, 324 in urban black households and 413 in urban coloured households.

Approximately 79% of households included in the 2009 survey were the same ones that participated in the earlier 2002 wave. A significantly higher proportion of rural black (94%) households than urban black (72%) and urban coloured (71%) ones were traced. A household that could not be traced was replaced by another older household in the same enumerator area. An estimated 69% of the 4199 household members enumerated in 2002 were traced to 2009. In total, 1286 individuals could not be traced. In this group 18% were reportedly temporarily absent, 55% had moved away permanently, and 27% (or 346 individuals) had died. This paper is based on information supplied by a total of 1059 households in the 2009 survey: 362 rural black households, 299 urban black households, and 398 urban coloured households.

Brazil:
Note that some of the information on sampling for the following section was taken from a document originally written in Portuguese and translated using Google translate. The original document is available with this dataset and is titled: "Benefícios Não-Contributivos e o Combate à Pobreza de Idosos no Brasil"

The approach taken in Brazil was similar to the one taken in South Africa, as the territorial expansiveness made it difficult to obtain a nationally representative sample of with a relatively small number of households. The alternative was to seek to expand the regional coverage as far as possible within the research budget. Two large regions were selected for field research. The first was the metropolitan area of Rio de Janeiro, in which the population of Rio de Janeiro state is most heavily concentrated. This is one of the most developed states in the country. Four counties were chosen within the metropolitan area. Three neighboring counties, Duke Caxias, Nova Iguaçu and São João de Meriti, were also selected. To represent the elderly population of the poorest regions of the country, a state in the Northeast was selected. Three possibilities were considered: Bahia, Pernambuco and Ceara. These have the the largest populations in the Northeast. The state of Bahia was chosen because of its proximity to Rio de Janeiro (making it more affordable to process the data). Of the major cities of Bahia, Ilheus was chosen as it had a more rural population, which the study aimed to capture.

The sample target was defined at around a thousand households with at least one person aged 60 or over in the household. Aiming to diversifying the population surveyed, the sample was divided into four groups, each with about one fourth of the sample. Thus, the state of Rio de January was half of the sample, and the rest distributed in the three counties in the Rio de Janeiro metropolitan area. The other half was divided in two, half being in the urban, and the other rural, in the municipality of Ilheus.

To select of households within each municipality the Brazilian 2000 Census data was used. Sectors with low income and high population of elderly, maximizing the probability of finding elderly not receiving contributory benefits, were chosen. The criteria used were:

1. At least 100 homes in the sector
2. At least 60% of households whose income was, at most, equal to two times the minimum wage
3. A minimum of 8% of elderly (60 or older) in the population.

Sectors fulfilling the above criteria were chosen randomly within each sector. As a way to diversify the selection of households, a target was set between 16 and 20 households with elderly members (at least one aged 60 years or more). Researchers surveyed the chosen sectors until the quota was fulfilled. The sectors randomly selected could not always be surveyed, however, given the level of urban violence that hit Rio de January in the year of initial data collection. Territorial disputes between drug gangs prevented access to some sectors. These were replaced by sectors reserved for such eventualities. Given the fact that the majority of pensions in Brazil are contributory, randomly surveying households meant that there was the possibility of only obtaining a very small number people with non-contributory pensions. Thus, the selected households were supplemented with records from the beneficiaries of non-contributory pensions in their municipalities. The records, however, were very inaccurate, including deceased beneficiaries, non-existent addresses and many other problems. Despite the difficulties, the existence of the records is allowed the collection of a reasonable number of recipients of pensions and non-contributory retirement funds.

Upon returning in 2008, the Brazilian fieldwork team managed to discover 340 of the Rio de Janeiro families and 303 of the Ilheus families, a success rate of 67.3% and 60.4% respectively. 643 out of the original 1006 households were discovered in the same household. Beyond that, 28 additional families interviewed in 2002 were discovered in a different household within the same area. This meant that 165 replacement households were required in the Rio de Janeiro area and 199 in Ilheus. 363 new families, therefore, are included in the 2008 sample.

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

Face-to-face [f2f]

{'Abstract': '《老龄化、福祉与发展项目（Brazza2）》旨在探究巴西与南非受助家庭中，补助金、社会养老金等对贫困和脆弱性的影响。该调查旨在协助研究人员探讨社会援助在提升生活质量方面的程度，以及来自老年养老金和其他社会补助金的收入是否增强了社会养老金领取者和家庭成员的物质与主观福祉。研究还调查了关于幸运与不幸的看法，以提供线索说明社会援助在增强贫困家庭韧性及其独立于国家方面的作用。', 'Analysis unit': '家庭和个人', 'Universe': '南非：调查覆盖了农村东开普省黑人家庭以及开普敦都市西部省的黑人和有色人种家庭。', 'Kind of data': '样本调查数据', 'Sampling procedure': "南非：在南非，Development Research Africa 公司被委托进行数据收集。为了进行抽样，他们请求从Stats SA获取满足以下标准的EAs列表：
1. 主要为黑人或有色人种EAs
2. 主要由Stats SA定义为都市（正式或非正式）的西部省
3. 主要由Statssa定义为部落或半都市的东部省；且
4. 不包含机构或农业区域（这些EAs被排除在外）

这些CEAs被发送到DRA，分别在不同的Excel表格中，以下为每个司法管辖区下的标题：
1. 按家庭户主人口群体划分的地理区域（非洲/黑人或有色人种），按人口加权
2. 按人口普查区域类型划分的地理区域（农村：部落村庄，都市：正式或非正式），按人口加权
3. 按年龄划分的地理区域（56岁及以上），按人口加权
4. 按家庭划分的地理区域（提供了每个CEA的家庭总数）

这些数据文件被整理并合并成三个单独的表格，反映了受访者类别。所有包含少于八十户家庭的CEAs都被删除，以进一步确保机构或农业区域（以及东部省的都市区域）不会成为合格对象，并且也限制了选择没有合格受访者家庭的CEAs的可能性。这三个数据库成为选择样本的三个样本框。

所有剩余的CEAs按升序排列。使用PSM抽样方法选择样本。这意味着拥有更多家庭数量的CEAs有更大的机会被选入样本。在选定的EAs下方直接的两个CEAs被包括为可能的替代品。一旦选定了EA编号，就从Stats SA获取地图。只有在这种情况下，才能确定这些CEAs的位置。由于PSM方法，较小司法管辖区的小EAs未能被选入样本，而较大司法管辖区则选入了多个EAs。在开普敦，EAs在开普敦街道地图上可以相对容易地找到。

每个受访者类别选择了二十个集群或EAs。每个类别的目标是大约333次访谈。因此，每个CEA需要进行大约17次访谈（333/20=17）。在集群或EA中需要接触的家庭数量是分段大小。分段大小取决于包含至少一名55岁及以上人员的家庭百分比以及假设的响应率。样本中每个CEA的分段大小都是单独计算的。例如，如果CEA中有120户家庭，居住着33名55岁或以上的居民，并且假设95%的响应率，那么在CEA中需要接触59户家庭（17/(15/120)*0.95），以便在每个CEA中获得17次成功的访谈。本研究的一个局限性是，该公式没有考虑同一年龄类别中可能居住在家庭中的两个人或更多人。

一旦从Stats SA获得地图，现场工作人员通过识别EA边界并输入任何地图上的特征或更改来验证和更新地图。然后统计家庭数量，并将它们分成约等数量的段。通过将分段大小（如上段所述）除以在EA中找到并记录的实际家庭数量来计算分段数量。一些EA可能只有一个段（如果分段大小大于EA中的家庭总数），也可能有五六个段。然后随机选择一个段。在特定段中接触所有家庭，并识别和调查所有目标家庭。最后，在家庭内部，选择对家庭开支情况最了解的人作为首位受访者。之后，对所有55岁及以上的人进行访谈。现场工作人员必须对每个家庭进行三次访问，以最大限度地提高访谈完成的可能性。项目负责人监督完成访谈的数量。在似乎不太可能实现每个受访者类别区域333次访谈的整体目标的情况下，现场工作人员必须对整个EA进行调查。

在农村东开普省随机选择的二十个EAs位于Zwelitsha、Keiskammahoek、Engcobo、Idutywa、Kentani、Libode、Lusikisiki、Mqanduli、Ngquleni、Nqamakwe、Port St Johns、Qumbu、Cofimvaba、Tabankulu、Tsomo、Willowvale和Lady Frere的司法管辖区。在开普敦都市的黑人家庭中随机选择的二十个EAs位于Goodwood、Wynberg、Mitchell's Plain（包括广阔的城镇Khayelitsha）和Kuils River的司法管辖区。针对城市有色人种家庭的二十个随机选择的EAs位于开普敦都市的同一司法管辖区，增加了Bellville。

2002年的样本设计规定，在最后阶段，在EA分段中选出的所有家庭都必须进行访谈。因此，2002年实现了比原计划的1000次访谈更大的样本量。2002年共完成了1111次访谈：农村黑人家庭374次，城市黑人家庭324次，城市有色人种家庭413次。

在2009年的调查中，约79%的家庭与早期2002波次中参与的家庭相同。农村黑人家庭（94%）比城市黑人家庭（72%）和城市有色人种家庭（71%）的比例显著更高。在2002年记录的4199户家庭成员中，估计有69%在2009年被追踪到。总共有1286人无法追踪。在这组人中，18%据报道暂时缺席，55%永久搬迁，27%（或346人）已去世。本文基于2009年调查中1059户家庭提供的信息：362户农村黑人家庭，299户城市黑人家庭，398户城市有色人种家庭。

巴西：
注意，以下部分关于抽样的信息部分取自原始用葡萄牙语撰写的文件，并使用Google翻译进行翻译。原始文件与该数据集一起提供，标题为：“巴西非贡献性福利与老年人贫困的斗争”。

在巴西采取的方法与南非类似，因为领土辽阔，难以获得由相对较少家庭组成的全国代表性样本。替代方案是在研究预算范围内尽可能扩大区域覆盖范围。选择了两个大型区域进行实地研究。第一个是里约热内卢大都市区，里约热内卢州的居民最为集中。这是该国最发达的州之一。在大都市区选择了四个县。选择了三个相邻的县，杜克·卡西亚斯、诺瓦伊瓜苏和圣若昂·德梅里蒂。为了代表国家最贫困地区的老年人口，选择了一个东北部的州。考虑了三个可能性：巴伊亚、佩尔南布科和塞阿拉。这些是东北部人口最多的州。由于与里约热内卢的邻近（使其处理数据更具成本效益），选择了巴伊亚州。在巴伊亚的主要城市中，伊卢伊斯被选中，因为它有更多农村人口，而该研究旨在捕捉这一点。

样本目标是定义在大约一千户家庭中，至少有一名60岁或以上的人。为了使调查的人口多样化，样本分为四组，每组约占样本的四分之一。因此，里约热内卢州是样本的一半，其余分布在里约热内卢大都市区的三个县。另一半分为两部分，一半在城市，另一半在农村，在伊卢伊斯市。

为了在每个市政区域内选择家庭，使用了巴西2000年人口普查数据。选择了低收入和高老年人口比率地区，最大程度地增加了找到未接受贡献性福利的老年人的可能性。使用的标准是：
1. 该区域至少有100户人家
2. 至少60%的家庭的收入最多等于最低工资的两倍
3. 人口中至少有8%的老年人（60岁或以上）

符合上述标准的地区在每个地区随机选择。为了使家庭选择的多样性，设定了16至20户有老年成员（至少一名60岁或以上）的目标。研究人员调查了选定的地区，直到达到配额。随机选择的地区可能无法进行调查，因为里约热内卢在数据收集当年的城市暴力水平。毒品帮派之间的领土争端阻止了对某些地区的访问。这些地区被预留用于此类情况。鉴于巴西的大多数养老金都是贡献性的，随机调查家庭意味着可能只获得非常少数的非贡献性养老金的人。因此，选定的家庭被其市政区域内非贡献性养老金领取者的记录补充。然而，这些记录非常不准确，包括已故领取者、不存在的地址和其他许多问题。尽管困难重重，但这些记录的存在使得收集了一定数量的养老金和非贡献性退休基金领取者。

2008年返回时，巴西实地研究团队设法发现了340户里约热内卢家庭和303户伊卢伊斯家庭，成功率分别为67.3%和60.4%。2002年在同一地区发现的28户额外家庭在2008年被发现。这意味着需要165户替换家庭在里约热内卢地区，以及199户在伊卢伊斯。因此，2008年样本中增加了363个新家庭。

数据收集方式：面对面（f2f）"}