Hungry Cities Partnership Survey, Nairobi 2016-2019 - Kenya

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

This study covers Nairobi, one of four African cities surved between 2013 and 2019 by the African Center for Cities. The African Center for cities is based at the University of Cape Town and is a partner of the Hungry Cities Partnership (HCP).

The HCP studies include household data on food insecurity, household food purchasing dynamics, nutritional discounting taking place in households, foods consumed and multidimensional measures of poverty. The household data is complimented with household member data and food retailer (vendor) data, including infomation on vendor employees.

The Hungry Cities Partnership is an international network of cities and city-based partner organizations which focuses on the relationships between rapid urbanization, informality, inclusive growth and urban food systems in the Global South.

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

The household sample aims to be representative of the city of Nairobi.

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

Households and individuals

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

Households and Vendors in Nairobi.

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

Sample survey data

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

Household: The survey report explains a two-stage sampling process. Firstly, 23 of the 111 administrative "sub-locations" in Nairobi were selected. 3 were selected randomly from each of the 8 divisions, excluding Kasarani which only had 2 selected. Then for the second stage, households were selected randomly from these 23 administrative locations. The number of households selected was proportional to the size (measured in households) of the administrative location, a "portional-to-population" strategy that was also employed in the Maputo household survey from 2014.

Vendor: 1267 food vendors were interviewed across Nairobi. The documentation sounds like a two-stage process was also followed. It states that at least 3 neighbourhoods were randomly selected in each of the eight administrative divisoins. The documentation states that vendors were randomly selected in the second stage, but then adds the following "respondents were randomly selected depending on the form and density of the location of business enterprises in the residential neighbourhood, category of food vendor, types of food items sold by the vendor, as well as willingness to participate in the survey." It is as such not clear to what extent this process was random.

Sampling deviation
---------------------------

Due to security reasons, the survey could only take place during daylight hours and as such the study may have missed businesses operating solely at night.

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

Face-to-face [f2f]

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

There are two questionnaires per city, a household questionnaire and a vendor questionnaire. The household questionnaire has a subsection for household members (persons), and the vendor quesitonnaire has a subsection for employees. Answers to these subsections are supplied in separete datafiles, which can be matched to (merged with) the questoinnaire as necessary.

DataFirst has not received documentation to confirm this, but it is likely that the protocols from the other cities were followed, in that 1) vendor surveys were administered to the person directly responsible for the running of the business using handheld tablets and 2) the household survey was administered to a senior adult member of the household, someone who could speak for the household.

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

Datafiles were received by DataFirst in SPSS (.sav) and Excel (.xlsx) format. Variables had to be named and variable labels were taken from question text. Variables were named accoriding to question number and subject matter, in a hierachical fasion.

An effort was made to keep question numbers and value labels consistent across cities where the same questions were asked for the 2013-2019 surveys. For the vendor data, Cape Town, Maputo and Nairobi had almost identical questionnaires and so the question numbers were naturally the same across these cities (harmonized). For the household data, Maputo, Nairobi and Windhoek were similar and could be harmonized. This means users could try stack these datafiles. This also means that list numbers/value codes might not match the questionnaire for a given city.

Missing values of 97, 98, and 99 were converted to -97, -98 and -99. There were some question numbers wrong in the vendor data questionnaires (typos) that were corrected.

In the household data, the confusingly numbered 10.c and 10.d were renamed to 10b1 and 10b2, to avoid confusion with 10c. and 10d., which were different questions.

Data appraisal
---------------------------

In general the lists change subtly between cities, for example the lists of foods in question 8 of the household data. As such the user should take caution when comparing across cities, and refer to the questionnaires. When the lists differed in a potentially confusing way, list item letters (a-z) were left in the variable name as a second way for the user to check that the data match the questionnaire correctly.

{'Abstract': '本研究涵盖内罗毕，该城市为非洲四座被非洲城市中心（非洲城市中心）在2013年至2019年间调查的城市之一。非洲城市中心位于开普敦大学，是饥饿城市伙伴关系（HCP）的合作伙伴。

HCP的研究包括关于食物不安全、家庭食品购买动态、家庭内发生的营养折扣、消费食品以及贫困的多维度衡量等方面的家庭数据。家庭数据辅以家庭成员数据和食品零售商（摊贩）数据，包括关于摊贩员工的信息。

饥饿城市伙伴关系是一个国际城市和城市合作伙伴组织网络，专注于全球南方快速城市化、非正式性、包容性增长与城市食品系统之间的关系。

地理覆盖范围
---------------------------
家庭样本旨在代表内罗毕市。

分析单元
---------------------------
家庭和个人

总体
---------------------------
内罗毕的家庭和摊贩。

数据类型
---------------------------
样本调查数据

抽样程序
---------------------------
家庭：调查报告解释了一个两阶段抽样过程。首先，从内罗毕111个行政“次地区”中选择了23个。每个8个区中随机选择了3个，排除了只有2个被选中的卡萨拉尼。然后在第二阶段，从这些23个行政地点中随机选择了家庭。所选家庭的数量与行政地点的大小（以家庭数量衡量）成比例，这是一种在2014年的马普托家庭调查中也采用的“按人口比例”策略。

摊贩：在内罗毕对1267名食品摊贩进行了访谈。文件表明，也遵循了两阶段过程。它表示在每个八个行政区分中至少随机选择了3个邻里。文件表示，在第二阶段随机选择了摊贩，但随后补充说“受访者根据居住区商业企业所在地点的形式和密度、食品摊贩类别、摊贩销售的食品类型以及参与调查的意愿随机选择。”因此，这一过程有多随机尚不清楚。

抽样偏差
---------------------------
由于安全原因，调查只能在白天进行，因此研究可能遗漏了仅夜间营业的企业。

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

研究工具
---------------------------
每个城市有两个问卷，一个是家庭问卷，另一个是摊贩问卷。家庭问卷有一个关于家庭成员（个人）的子部分，摊贩问卷有一个关于员工的部分。这些子部分的答案提供在单独的数据文件中，可以根据需要与问卷（合并）匹配。

DataFirst尚未收到确认此点的文件，但很可能遵循了其他城市的协议，即1）对直接负责经营业务的人进行摊贩调查，使用手持平板电脑；2）对家庭进行调查的是家庭中的年长成员，他能代表家庭。

数据清理操作
---------------------------
数据文件以SPSS (.sav)和Excel (.xlsx)格式收到。必须命名变量，并将变量标签从问题文本中提取。变量名称根据问题编号和主题进行命名，采用分层方式。

努力保持问题编号和值标签在城市之间的一致性，其中相同的2013-2019调查问题被提出。对于摊贩数据，开普敦、马普托和内罗毕的问题问卷几乎相同，因此这些城市的问卷编号自然相同（协调一致）。对于家庭数据，马普托、内罗毕和温得和克相似，可以协调一致。这意味着用户可以尝试堆叠这些数据文件。这也意味着列表编号/值代码可能与给定城市的问卷不匹配。

将97、98和99的缺失值转换为-97、-98和-99。在摊贩数据问卷中存在一些问题编号错误（拼写错误），已予以纠正。

在家庭数据中，混乱编号的10.c和10.d被重命名为10b1和10b2，以避免与10c.和10d.（不同问题）混淆。

数据评估
---------------------------
一般来说，列表在不同城市之间变化微妙，例如家庭数据问题8中的食品列表。因此，用户在比较城市时应谨慎，并参考问卷。当列表以可能令人困惑的方式不同时，列表项字母（a-z）被留在变量名中，作为用户检查数据是否正确匹配问卷的第二种方式。'}