Census of Agriculture 2008-2009 - Uganda

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

The agricultural sector is the most important sector of the Ugandan economy. Empirical evidence attests to this; for example the share of the agricultural sector to Gross Domestic Product (GDP) is about 21 percent (at the then current prices). According to the Agricultural Module of the 2002 Population and Housing Census, the agricultural sector accounted for 73 percent of the total employment for the persons aged 10 years and above. In addition, 74 percent of the households had an agricultural holding. The long term vision of the Government of Uganda is to eradicate poverty and the strategies for this vision are defined in the then Poverty Eradication Action Plan (PEAP) which has been transformed into the National Development Plan (NDP).

The vision of PMA was to eradicate poverty through transforming subsistence agriculture to commercial agriculture. The whole process of transformation requires accurate and reliable agricultural data to monitor the progress made and inform policy and planning processes

Further, countries are focusing on the need to monitor progress towards the Millennium Development Goals (MDGs) through their National Statistical systems. The World Census of Agriculture (WCA), 2010 was formulated with this in mind and specifically to monitor eradication of extreme poverty and hunger, achievement of Universal Primary Education, Promotion of gender equality and empowerment of women and ensuring environmental sustainability.

Within the framework of the FAO/World Bank Agricultural Statistics Assistance to Uganda, a Data Needs Assessment Study was undertaken in August 1999. One of the major findings was that the Agricultural Statistics System was fragile, vulnerable, un-sustainable and above all, unable to meet the data needs of users. A Census of Agriculture (CA) is major source to meet these demands.

Census taking in Uganda
Prior to the conducting of the Uganda Census of Agriculture (UCA), 2008/09 two (2) other censuses had been conducted. The first CA was conducted during 1963/65. The Government of Uganda was assisted by FAO and the then Department for Technical Cooperation of the United Kingdom both of which provided international and census equipment to a varying degree.

The second CA called the National Census of Agriculture and Livestock (NCAL) was conducted during 1990/91. It was funded by United Nations Development Programme (UNDP) and executed by FAO. Therefore the UCA 2008/09 formed the third CA in the history of census taking in Uganda.

Preparatory activities
An Agricultural Module was included in the Population and Housing Census 2002, to collect the data that would form a basis for constructing an up-to-date and appropriate sampling frame for a Uganda Census of Agriculture (UCA), 2004/05. A Pre-Test was conducted in 2002 followed by a pilot Census of Agriculture (PCA) which was conducted in 2003.

Lack of financial resources militated against conducting the UCA, 2004/05. During the Financial Year (FY) 2007/08 Government made a budgetary provision for conducting a census of agriculture.

The FY 2007/08 was mainly a preparatory year. As mentioned earlier, the plan had been to conduct a UCA during 2004/05, which did not take place. By 2008/09 (the census reference year), many changes had taken place and needed to be addressed. To this end, another Pre -Test was conducted in May 2008. Based on the findings from the Pre-Test, the UCA instruments had to be revised. Another very important factor for the instruments' revision was an input from the International Consultants (like FAO Statisticians).
Other preparatory activities included arrangements to procure census equipment and transport as well as recruiting and training of Field Staff.

Objectives of the UCA.2008/09
While the long-term objective of the UCA, 2008/09 was to have a system of Food and Agriculture Statistics (FAS) in place, the immediate objective was to collect and generate benchmark data needed for monitoring and evaluation of the agricultural sector at all levels, through a nation-wide CA.

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

The Uganda Census of Agriculture 2008/09 covered all the 80 districts in the country as of July 2007.

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

Agricultural households, Agricultural holdings

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

The Uganda Census of Agriculture 2008/09 was therefore planned to cover all the 80 districts at the time and collect data on various structural characteristics of agricultural holdings. Limited data on livestock variables was planned to be collected because comprehensive livestock data was to be collected in a Livestock Census, 2008.

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

Census/enumeration data [cen]

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

A stratified two-stage sample design was used for the small and medium-scale household-based agricultural holdings. At the first stage Enumeration Areas (EAs) were selected with Probability Proportional to Size (PPS), and at the second stage, households which were the ultimate sampling units were selected using systematic sampling.

For each of the sampled EAs, listing took place in the field and a number of filter questions (using Listing Module) were administered to determine eligibility (i.e., only the Households with Agricultural Activity would be eligible). Further, the eligible households were stratified into two strata namely, the small/medium holdings stratum and the Private Large-Scale holdings stratum.

On the other hand, district supervisors compiled separate lists of Institutional Farms and Private Large Scale Farms. These were to be covered on a complete enumeration basis.

During sampling, two (2) lists namely for EAs and PLS&IFs were used to identify possibilities of duplication and address them. If a PLS&IF was in both lists, it was deleted from the EA frame. However, if it was found only in the EA frame, it was left as part of the frame from which to sample. In other words, the List was not updated based on the information collected from the EAs sampled from the Area Frame.

The UCA2008/09 estimates were planned to be generated at national, regional and district levels. To achieve this, a sampling scheme of 3,606 EAs and 10 agricultural households in each selected EA, leading to 36,060 households was adopted.

In this design, an optimum number of households to be sampled per EA was determined on the basis of a suitable cost ratio (ratio of the cost per PSU to cost per SSU) and intra-class correlation, calculated from the Agricultural Module data from PHC 2002. For a cost ratio of 40 and intra-class correlation as 0.29, optimum number of households to be selected was obtained as 10.

The required sample size of EAs was selected from each district with probabilities proportional to size (PPS), using the systematic sampling algorithm described in Hansen, Hurwitz, and Madow (1953) while Agricultural Households were selected with equal probability systematic sampling procedure. The measure of Size (MOS) which was used for sample selection was the number of Agricultural Households determined from the 2002 PHC.

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

EAs where there was no enumerations due to insecurity: There were EAs which could not be listed or even enumerated due to insecurity , resistance by residents or nonexistent etc. These were in Moroto, Nakapiririt, Mubende, Kampala etc. Since there were no replicate EAs, the number of sampled EAs in those districts was lowered reducing the estimated number of EAs expected to give good results in those respective districts.

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

Face-to-face [f2f]

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

The principles of validity, optimization and efficiency which refer to ability for the questionnaires to yield more reliable information per unit cost; measured as a reciprocal of the variance of the estimate and enables objective interpretation of the results was followed. While costs involved man hours and money expended for data collection from sampled units, the design of questionnaires had to collect a minimum set of internationally comparable core data(indices) for Uganda, as enshrined in the pillars of FAO.

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

Data Processing monitored the data quality parameters and data quality team could continuously report to the field operations team who could make feed back to the DSs for improvement. Returned questionnaires were subjected to the following steps Coding, Data capture, Editing, Secondary Editing and Quality control.

Coding
This involved making sure that all forms/questionnaires had correct geographical identification information and correct crop codes. The coding team reviewed the sampling of holdings within an enumeration area to see that only eligible/sampled holdings were actually enumerated.

Editing
This involved the process of identifying inconsistencies within the data and removing them. At the beginning of UCA data processing, a set of editing rules and guidelines where developed by the data processing team with technical guidance from the subject matter specialists. Many of these were incorporated into the data entry application and others were left for the secondary editing stage.

Secondary Editing
Errors that passed the data entry stage were subjected to the editing stage. This stage was meant to find inconsistencies within the data. It brought out problems that required subject matter specialists to resolve. To resolve most of such errors, consultations were made with the national supervisors, district supervisors, UBOS and MAAIF technical teams.

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

The UCA2008/9 had several forms namely; Agricultural Households and holding Characteristics Module; Crop Area Module; Crop Production Module and PLS & IFs Module. Overall, the response rate was 93.5 percent.

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

The estimated number of Ag HHs was 3.95 million. Out of this, the Western Region had the highest percentage (28.5 percent) closely followed by the Eastern Region with 28.1 percent while the Northern and the Central Region had the 22.9 and 20.5 percent respectively.

Out of about 3.6 million Ag HHs with information on the sex of HH head, 2.8 million (78.9%) and 754,000 (21.1%) were Male and Female respectively

摘要
---------------------------
乌干达农业部门是该国最重要的经济部门。实证证据证实了这一点；例如，农业部门占国内生产总值（GDP）的比重约为21%（按当时的价格计算）。根据2002年人口和住房普查的农业模块，农业部门占10岁及以上人口的就业总量的73%。此外，74%的家庭拥有农业资产。乌干达政府的长期愿景是消除贫困，而实现这一愿景的战略则在当时的消除贫困行动计划（PEAP）中得以界定，该计划已演变为国家发展计划（NDP）。

PMA的愿景是通过将自给自足的农业转变为商业农业来消除贫困。这一转变过程需要准确可靠的数据来监测进展情况，并指导政策和规划过程。

此外，各国正专注于通过其国家统计系统监测实现千年发展目标（MDGs）的进展。2010年世界农业普查（WCA）正是基于这一考虑而制定的，其具体目标是监测极端贫困和饥饿的消除、普及小学教育、促进性别平等和赋权女性，以及确保环境可持续性。

在FAO/世界银行向乌干达提供农业统计援助的框架下，于1999年8月开展了一项数据需求评估研究。其中一项主要发现是，农业统计系统脆弱、易受攻击、不可持续，最重要的是，无法满足用户的数据需求。农业普查（CA）是满足这些需求的主要来源。

乌干达的农业普查
---------------------------
在2008/09年的乌干达农业普查（UCA）之前，已经进行了两次普查。第一次农业普查于1963/65年进行。乌干达政府得到了FAO和当时英国技术合作部的援助，两者在提供国际和普查设备方面程度不同。

第二次农业普查被称为国家农业和畜牧业普查（NCAL），于1990/91年进行。该普查由联合国开发计划署（UNDP）资助，由FAO执行。因此，UCA 2008/09成为了乌干达普查历史上的第三次农业普查。

准备工作
---------------------------
在2002年人口和住房普查中包含了农业模块，以收集构成乌干达农业普查（UCA）2004/05年最新和适当的抽样框架的数据。2002年进行了一次预测试，随后于2003年进行了一次试点农业普查。

由于缺乏资金，UCA 2004/05年未能进行。在2007/08财年，政府为进行农业普查做出了预算安排。

2007/08财年主要是准备年。如前所述，原计划于2004/05年进行UCA，但未能实现。到2008/09年（普查参考年），许多情况发生了变化，需要解决。为此，2008年5月又进行了一次预测试。根据预测试的结果，需要对UCA工具进行修订。修订工具的另一个非常重要的因素是国际顾问（如FAO统计学家）的反馈。

其他准备工作包括采购普查设备和运输工具、招募和培训现场工作人员等。

UCA.2008/09的目标
---------------------------
尽管UCA 2008/09的长期目标是建立食品和农业统计（FAS）系统，但近期目标是收集和生成用于监测和评估农业部门所有层次的数据，通过全国性的农业普查来实现。

地理覆盖范围
---------------------------
乌干达农业普查2008/09涵盖了截至2007年7月的所有80个区。

分析单元
---------------------------
农业家庭，农业资产

总体
---------------------------
因此，乌干达农业普查2008/09计划涵盖当时所有的80个区，并收集有关农业资产的各种结构特征数据。由于全面收集畜牧业数据将在2008年的畜牧业普查中进行，因此计划仅收集有限的畜牧业数据。

数据类型
---------------------------
普查/登记数据 [cen]

抽样程序
---------------------------
对于小型和中型家庭为基础的农业资产，使用了分层两阶段样本设计。在第一阶段，使用按规模成比例的概率（PPS）选择了普查区（EA），在第二阶段，使用系统抽样从家庭中选择最终的抽样单元。

对于每个样本EA，在实地进行登记，并使用登记模块中的若干过滤问题来确定资格（即，只有具有农业活动的家庭才有资格）。此外，合格的家庭被分为两个层次，即小/中规模资产层次和私人大型规模资产层次。

另一方面，区监督员编制了机构农场和私人大型农场各自的单独名单。这些将被完全普查的基础上进行。

在抽样过程中，使用了两个列表，即EA列表和PLS&IFs列表，以确定重复的可能性并解决这些问题。如果PLS&IF在两个列表中，则从EA框架中删除。然而，如果仅在EA框架中发现，则将其保留为从该框架中进行抽样的部分。换句话说，列表没有根据从区域框架中抽样的EA收集的信息进行更新。

UCA2008/09的估计计划将在国家、地区和区级生成。为此，采用了一个包含3,606个EA和每个选定EA中的10个农业家庭的抽样方案，导致36,060个家庭。

在这个设计中，每个EA的抽样家庭数量是在一个合适的成本比率（每PSU成本与每SSU成本之比）和来自2002年PHC的农业模块数据的类内相关性的基础上确定的。对于成本比率为40和类内相关系数为0.29，最佳抽样家庭数量为10。

从每个区选择了所需数量的EA样本，使用与规模成比例的概率（PPS），并使用Hansen，Hurwitz和Madow（1953年）描述的系统抽样算法。农业家庭使用等概率系统抽样程序进行选择。用于样本选择的规模（MOS）措施是从2002年PHC确定的农业家庭数量。

抽样偏差
---------------------------
由于不安全而未能进行普查的EA：由于不安全、居民抵制或不存在等原因，有些EA无法进行登记或甚至进行普查。这些地区包括Moroto，Nakapiririt，Mubende，Kampala等。由于没有重复的EA，这些区的样本EA数量减少，导致预计在这些相应地区将给出良好结果的EA估计数量减少。

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

研究工具
---------------------------
遵循了有效性、优化和效率的原则，这些原则指的是调查问卷以每单位成本产生更可靠信息的能力；作为估计方差的反比，并使结果具有客观的解释能力。虽然涉及的人力物力成本和为从样本单元收集数据而支出的资金，但问卷设计必须收集乌干达作为FAO支柱所体现的最低国际可比核心数据（指数）。

数据清理操作
---------------------------
数据处理监测了数据质量参数，数据质量团队可以持续向现场运营团队报告，现场运营团队可以对DSs提出反馈以进行改进。退回的问卷将进行以下步骤：编码、数据录入、编辑、二级编辑和质量控制。

编码
这涉及到确保所有表格/问卷都具有正确的地理识别信息以及正确的作物代码。编码团队审查了普查区内的持有抽样，以确保只有合格/抽样的持有实际上被普查。

编辑
这涉及到识别数据中的不一致并删除它们的过程。在UCA数据处理的开始阶段，数据处理团队开发了一套编辑规则和指南，并从主题专家的技术指导中获得了指导。其中许多被纳入数据录入应用程序中，其他则留给了二级编辑阶段。

二级编辑
通过数据录入阶段的数据将受到编辑阶段的审查。这一阶段旨在发现数据中的不一致。它揭示了需要主题专家解决的问题。为了解决此类错误中的大多数，与国家监督员、区监督员、UBOS和MAAIF技术团队进行了咨询。

响应率
---------------------------
UCA2008/9有几种形式，包括农业家庭和持有特征模块；作物面积模块；作物生产模块和PLS & IFs模块。总体而言，响应率为93.5%。

数据评估
---------------------------
估计的农业家庭数量为390万。其中，西部地区占比最高（28.5%），其次是东部地区（28.1%），北部和中部地区分别为22.9%和20.5%。在约360万个有家庭负责人性别信息的农业家庭中，280万（78.9%）为男性，75.4万（21.1%）为女性。