National Agricultural Census 2009 - Fiji

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

The Fiji National Agricultural Census 2009 is the fourth agricultural census. After a lapse of 18 years, National Agricultural Census 2009 was carried out in Fiji beginning in October 2009; data collection was interrupted by two cyclones and not completed until March of 2010.

The Agriculture Census is a national obligation conducted by the country to provide benchmark data for planning and policy decisions in sustainable agricultural and rural development; and to strengthen and improve the ongoing Fiji Agriculture Statistics System (FASS) to generate key agricultural data on a regular basis using the results of the 2009 NAC as the benchmark and the dissemination of this statistical information in the form of regular reports.

The 2009 National Agriculture Census (NAC) is the first census programme to be conducted in the country using Multiple Sample Frame (MSF) as the main methodology. Given the experiences of the previous census programmes in terms of funding and availability of resources, the 2009 agriculture census programme provides a platform for more diversification and improvement programmes within the agriculture sector thus ensuring compatible foreign exchange earnings as well as uplifting the living standards of rural populace.

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

National

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

- Agricultural Holding and Holders

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

Census/enumeration data [cen]

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

The survey design used the multiple sampling frame methodology. This methodology combines the advantages of an area frame (complete coverage) and a list frame (rare commodities and large and special farms). In the 2009 NAC, it was expected to provide reliable results at district level for most tables, although results for smaller districts might not be possible. In addition, a small island strategy (SIS) was used where complete enumeration of villages occurred within some districts.

The underlying basis for an area frame sample is to select small areas (in this case, one square kilometer - 100 hectares) that represent the entire area of interest. To improve the efficiency of the sample, the entire country was stratified (or characterized) by the intensity of agriculture. The stratification split the country into areas of high intensity agriculture, medium intensity agriculture, low intensity agriculture, forest areas, peri-urban areas and urban areas/non agricultural areas. The overall sample size was limited by the resources available; it was determined to use a ten percent sample of "agricultural land" as determined during the stratification process.

Initially the Fiji Bureau of Statistics (FIBOS) enumeration areas (EAs) for the 2007 Population and Housing Census were used for stratum identification. Subsequently it was determined that re- stratification of whole EAs and subdivision of other EAs would be more efficient. In many of the FIBOS EAs, farms were present only in small pockets; the uniformity of agriculture in the EA, one of the strengths of the stratification, did not exist. These EAs were, first, reviewed for the presence of natural pine forest and natural reserves. After these areas were removed, the remainder of the EA was divided into one square kilometer grids before the sampling process occurred. After the grids were selected, the Land Use Section of the DOA prepared maps using detectable boundaries "around the grid". It was not possible for segments to retain the gridlines as boundaries because they seldom were along recognizable boundaries; however, it was possible to approximate 100 hectares in that general area.

A farm can consist of land areas that are separated by physical boundaries or by land use patterns; these are called tracts. The method of data collection was to account for each tract inside the segment, but, also to collect information about areas outside the segment for farms with tracts both inside and outside. If a segment boundary splits an existing tract, it is divided into one tract inside the segment and one tract outside the segment. The percentage of the farmland inside the segment is used as a weighting factor for the farm in the expansions.

One of the limitations of area frame samples is the accurate expansion of rare or concentrated (non- uniform) variables - such as poultry houses or large dairy or beef farms. The list frame sample, developed from the knowledge and experience of DOA Animal Health and Production Division and Extension Division staff, was expanded as data collection occurred and there was better awareness of large and specialized farms. Data were collected from all of these farms. It should be noted that shortly before the beginning of data collection, a severe outbreak of brucellosis occurred and some culling took place.

Three levels of data presentation were identified for tabulation of the data of the National Agriculture Census 2009 (NAC 2009). The first is tables and expansions at district level; the second is tables and expansions at provincial and national level; the third is tables and (estimates) for special variables.

The census data were collected at farm level, at tract level, at crop level and at animal/poultry level. Information about households and their demographics were also collected. One priority area has been the role of gender in agriculture in Fiji. A special section of the census questionnaire was targeted at identifying these roles and highlighting any special differences. These data also have been broken out by age group.

Accurate land stratification for the 2009 NAC was essential; it was necessary to estimate the percentage of agriculture land use. Initially the stratification was made for each of the Fiji Islands Bureau of Statistics (FIBOS) enumeration areas (EAs).

The census estimates were requested at national, divisional, provincial and also tikina levels. The 15 provinces including Rotuma Island were the main focus of the tabulation. Consequently, the entire country was divided into strata according to the intensity of land use for agriculture. They were further subdivided into sub-strata according to specific land use. This sub-stratification technique guaranteed the sample allocation for priority and special crops. Another stratum was created for special farms including large commercial and freehold farms.

A total of 1,602 existing EAs from the 2007 population census were overlaid on the ASF topographic maps scale 1:50,000 in preparation for stratification activities according to land use. Each EA was classified into one of the strata keeping the same geographical identification codes as those used in the population census. The percentage of area under crops, pastures, forest, etc. (land use) of each EA was estimated by field observation to check that each EA was classified in the right stratum and sub-stratum.

The sampling procedures are more fully described in "National Agricultural Census 2009 - Final Report" pp.7-13.

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

Face-to-face [f2f]

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

Two questionnaires, NAC 1 and NAC 3, were used to record information about the segments from the sample. The NAC 1 itemized all tracts inside the segment and all associated farm tracts outside the segments. The NAC 3 documented the nonfarm tracts inside the segment. Enumerators were required to fill out these questionnaires; during the interview process the main questionnaire (NAC 2) was used. Neither the NAC 1 nor NAC 3 was necessary for List Frame farms.

The questionnaire was designed and tested by the staff of the Agricultural Statistics Unit and training manuals were prepared for supervisors and enumerators. A Pilot Census was carried out in several locations to evaluate the content and layout of the questionnaires and the completeness of the census documents. The questionnaire and training materials were updated as the result of the Pilot Census.

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

After a prioritized order of data collection from the provinces, the questionnaires were received at the Agricultural Statistics Unit in batches. Unique questionnaire numbers were assigned by the data processing administrator and recorded in a management system designed to prevent duplicate numbers and to coordinate the collection and processing of the three types of questionnaires. The questionnaire numbers consisted of province, district and a sequence number starting with an initial value assigned previously to each of the segments.

The editing and coding process for a total of 9,341 NAC 2 questionnaires containing farm data started in mid November 2009. Four persons managed the archives of census materials (questionnaires, cartography and photo-enlargements, etc.). Eleven coders were contracted and trained using the Field Team Manual and the Coding, Editing and Data Processing Manual. One table head checked the manual editing and coding. Data entry activities were conducted by ten data entry operators beginning in early December.

Consistency checks were also carried out in the ACCESS databases. Queries were designed to identify data entry and coding errors. Data were entered into 15 provincial databases (including Rotuma Island) which were combined into four divisional databases. The LSF database was kept separate, but combined in SPSS for tabulation and analysis.

摘要
---------------------------
斐济国家农业普查2009年是第四次农业普查。经过18年的间隔，国家农业普查2009于2009年10月开始在斐济进行；由于两次飓风的干扰，数据收集工作直到2010年3月才完成。
农业普查是该国的一项国家义务，旨在为可持续农业和农村发展提供基准数据，以供规划和政策决策之用；同时，旨在加强和改进斐济农业统计系统（FASS），利用2009年国家农业普查（NAC）的结果作为基准，并通过定期报告的形式发布这些统计信息。
2009年国家农业普查（NAC）是该国首次采用多重样本框架（MSF）作为主要方法的普查项目。鉴于先前普查项目在资金和资源可用性方面的经验，2009年农业普查项目为农业部门的多样化和发展提供了平台，从而确保了外汇收入的兼容性，并提升了农村居民的生活水平。
地理覆盖范围
---------------------------
全国
分析单位
---------------------------
- 农业持有者和持有单位
数据类型
---------------------------
普查/清点数据 [cen]
抽样程序
---------------------------
调查设计采用了多重样本框架方法。该方法结合了区域框架（全面覆盖）和名单框架（稀有商品和大型特殊农场）的优点。在2009年NAC中，预计将提供可靠的结果，用于大多数表格的区级分析，尽管对于较小地区的分析可能无法实现。此外，在部分区使用小岛策略（SIS），在这些区内对村庄进行了全面清点。
区域框架样本的基础是在代表整个感兴趣区域的小区域（在本例中为1平方公里——100公顷）中进行选择。为了提高样本效率，整个国家根据农业强度进行了分层（或特征化），将国家划分为高农业强度区、中农业强度区、低农业强度区、森林区、近郊区和城市区/非农业区。整体样本量受可用资源的限制；决定在分层过程中使用10%的“农业用地”样本。
最初，斐济统计局（FIBOS）2007年人口和住房普查的统计区域（EA）被用于分层识别。随后，决定对整个EA和其他EA进行再分层，这会更加高效。在许多FIBOS EA中，农场仅存在于小块区域；EA中农业的统一性，分层的一个优势，并不存在。首先，对这些EA进行了审查，以确定是否存在天然松林和自然保护区。在这些区域被移除后，EA的其余部分在抽样过程之前被划分为1平方公里的网格。在网格被选定后，DOA土地利用处的部门使用可检测的边界“围绕网格”绘制了地图。由于这些段很少沿着可识别的边界，因此无法保留网格线作为边界；然而，在该一般区域内可以近似100公顷。
一个农场可以由由物理边界或土地利用模式分隔的土地区域组成；这些被称为地块。数据收集的方法是计算每个段内地块的数据，同时也收集有关段外区域的数据，对于既有段内又有段外的地块。如果段边界分割了一个现有的地块，它将被划分为段内的一个地块和段外的一个地块。段内农田的百分比被用作该农田扩展的权重因子。
区域框架样本的一个局限性是难以准确扩展稀有或集中（非均匀）变量——如鸡舍或大型乳牛或牛肉农场。从DOA动物健康和生产部门以及推广部门工作人员的知识和经验中发展而来的名单框架样本，在数据收集过程中得到了扩展，并且对大型和专业农场有了更好的认识。收集了所有这些农场的数据。需要注意的是，在数据收集开始前不久，布鲁氏菌病爆发严重，进行了部分屠宰。
为2009年国家农业普查（NAC 2009）的数据编制确定了三个层次的数据展示。第一个是在区级层面进行的表格和扩展；第二个是在省和国家级别的表格和扩展；第三个是特殊变量的表格和（估计）。
普查数据是在农场水平、地块水平、作物水平和动物/家禽水平收集的。还收集了关于家庭及其人口统计信息。一个优先领域是斐济农业中的性别角色。普查问卷的一个专门部分旨在确定这些角色，并强调任何特殊差异。这些数据还按年龄组进行了分类。
对于2009年NAC的准确土地分层至关重要；需要估计农业用地使用百分比。最初，每个斐济群岛统计局（FIBOS）统计区域（EA）都进行了分层。
要求在国家、分区、省以及tikina级别进行普查估计。包括罗图马岛在内的15个省份是编制的重点。因此，整个国家根据农业用地强度划分为分层。它们进一步根据具体土地利用划分为亚分层。这种亚分层技术保证了优先和特殊作物的样本分配。为包括大型商业和自有农场在内的特殊农场创建了一个分层。
在准备根据土地利用进行分层活动的过程中，将2007年人口普查的1,602个现有EA叠加到ASF地形图的比例尺1:50,000上。每个EA被分类为一个分层，保持与人口普查中使用的相同的地理识别代码。
每个EA的作物、牧场、森林等（土地利用）面积百分比通过现场观察进行估计，以检查每个EA是否被正确分类到正确的分层和亚分层。
抽样程序在“2009年国家农业普查——最终报告”的第7-13页有更详细的描述。
数据收集方式
---------------------------
面对面 [f2f]
研究工具
---------------------------
使用了两个问卷，NAC 1和NAC 3，来记录样本段的信息。NAC 1列出了段内的所有地块和所有相关的外部农场地块。NAC 3记录了段内的非农场地块。调查员需要填写这些问卷；在访谈过程中使用了主要问卷（NAC 2）。NAC 1或NAC 3对于名单框架农场来说都不是必需的。
问卷由农业统计单位的员工设计和测试，并为监督员和调查员准备了培训手册。在几个地点进行了一次试点普查，以评估问卷的内容和布局以及普查文件的完整性。根据试点普查的结果，更新了问卷和培训材料。
数据清理操作
---------------------------
在从各省收集数据的过程中，问卷以批量形式提交到农业统计单位。数据处理管理员分配了唯一的问卷编号，并在旨在防止重复编号并协调三种类型问卷的收集和处理的系统中进行了记录。问卷编号由省、区和从先前分配给每个段的一个初始值开始的序列号组成。
从2009年11月中旬开始，对包含农场数据的9,341份NAC 2问卷的编辑和编码过程开始。雇佣了11名编码员，并使用现场团队手册和编码、编辑和数据处理手册进行了培训。一名表头检查了手册编辑和编码。从12月初开始，由10名数据录入员进行了数据录入活动。
还在ACCESS数据库中进行了一致性检查。设计了查询以识别数据录入和编码错误。数据被输入到包括罗图马岛在内的15个省级数据库中，这些数据库被合并成四个分区数据库。LSF数据库保持独立，但在SPSS中进行编制和统计分析。