Survey of Agricultural Holdings 2021 - Georgia

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

The sample design of the Production Methods and the Environment module survey is based on the sample of the current Survey of Agricultural Holdings, so firstly given the design of the current Survey. The main purpose of the Survey of Agricultural Holdings as well as Production Methods and the Environment module is to produce official indicators in line with agricultural sector. The survey allows the compilation of statistics on crops and animal husbandry, of which information annual and permanent crops, sown area, average yield of annual crops, farming practices and their linkages with the natural environment, crop and livestock production methods, access to and use of information services, infrastructure and communal resources and etc. Statistical tables are accessible through the following link: https://www.geostat.ge/en/modules/categories/196/agriculture.

Production Methods and the Environment Module is part of main Survey of Agricultural Holdings. One round of the main survey (reference year) includes 5 inquiries: The Inception interview is carried out using the inception questionnaire during the period of January-February of the reference year. During this interview the sampled holdings are identified and situation existing at the holding as of first January is recorded. I, II and III quarter interviews are conducted by means of quarterly questionnaire at the beginning of the following month of the corresponding quarter of the reference year. Based on these surveys, the information about agricultural activities during the corresponding quarter is collected. The final interview is conducted by means of final questionnaire in January of the following year of the reference year. During this interview, the information about agricultural activities at the holding during IV quarter of the reference year and the summery information about agricultural activities at the holding during the whole reference year (from 1 January to 31 December of the previous year) are collected. During all five interviews, the same agricultural holdings (about 12000) are interviewed which are selected by a two-stage stratified cluster random sampling procedure out of about 642 000 agricultural holdings operated in Georgia. On the first stage, clusters (settlements) are selected. On the second stage, holdings are selected within the selected clusters.

The survey completely covers the territory of Georgia, excluding the occupied territories of Autonomous Republic of Abkhazia and Tskhinvali region. Each year a new sample is selected based on a rotational design (on a 3-year basis). In particular, every year approximately 4000 holdings out of the 12000 sampled holdings are replaced by new holdings. Sampled holdings participate in the survey for 3 years. Large agricultural holdings are sampled every year with complete coverage. The statistical unit of the survey is the agricultural holding (family holdings and agricultural enterprises) - which is defined as an economic unit of agricultural production under single management comprising all livestock kept and all land used wholly or partly for agricultural production purposes, without regard to title, legal form or size. Agricultural activities are conducted under the supervision of a holder (in case of households - a member of household, in case of agricultural enterprises - director or authorized person), who is responsible for making decisions and takes all economic risks and expenses related to agricultural activities. More than 270 interviewers participate in the survey fieldwork. For the Data collection, computer-assisted personal interviewing method (CAPI) is used in the family holdings. In case of agricultural enterprises, the authorized persons of the enterprises (respondent) fill the electronic (online) questionnaires by themselves (CAWI). Coordination of the interviewers and the primary control of the collected data during the field is carried out by coordinators. Their working area covers several municipalities. The function of the coordinators also includes consultation for agricultural enterprises on methodological and technical issues related to the survey. Production Methods and Environment module field work was carried out from May 5th to May 20th of 2022. 200 field staff participated in the survey, 22 of which were field supervisors. In total 5,880 agricultural holdings were selected for the PME survey. Such are the extra-large farms that are continuously participating in the survey and the third rotation farms that have been participating in the survey since 2019. Currently 943 extra-large farms and 3,899 third rotation farms are participating in the survey. Therefore, we have a total of 4,842 farm data for the last three years. The rest of the holdings will be selected from the first rotation clusters where interviews have been conducted for two years. In particular, using simple random sampling approximately 30% of the working clusters of the first rotation are selected in each stratum. This will give us about 1,038 farms. A total of about 5,880 farms will be selected.

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

Entire country (Georgia), excluding occupied regions (Abkhazia and Tskhinvali region)

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

Agricultural holding – economic unit of agricultural production under single management comprising all livestock kept and all land used wholly or partly for agricultural production purposes, without regard to title, legal form or size in which agricultural activities are conducted under the supervision of a holder, who is responsible for making decisions and takes all economic risks and expenses related to agricultural activities.

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

Survey sampling frame includes about 642 000 agriculture holdings (households and agricultural enterprises) operated in country. The Agricultural Census 2014 is the main source of the sample frame. Sampling frame is updated on a permanent basis in according to the results of survey of agricultural holdings, business register and different administrative sources.

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

Sample survey data [ssd]

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

The sample design of the Production Methods and the Environment module survey is based on the sample of the current Survey of Agricultural Holdings, so firstly given the design of the current Survey.
• Main Source of the sample frame since 2016 - Agricultural Census 2014;
• Sample frame contained 642 000 holding - sample size 12 000 (1.9%);
• Sample Design: two-stage stratified cluster random sampling; - First stage - selection of cluster (Settlement); - Second stage - Selection of holdings within the selected clusters;
• Each year a new sample is selected based on a rotational design; - Every year 1/3 of holdings (4 000) selected a year before are replaced (Sampled holdings participate in the survey during 3 years);
• Extremely large agricultural holdings are sampled every year with complete coverage;
• Additional Sources for updating sample frame: Sample Survey of Agricultural Holdings, Statistical Business Register, Administrative data existing in MEPA (large agricultural holdings); Sampling error of main indicators do not exceed 5% for a country level and 10% for a regional level.

The sample design of the Production Methods and the Environment module survey:
• Sample Design: Two-stage cluster sampling was used for the survey. Sample is formed separately in each stratum. At first, clusters are selected in every stratum, and then holdings from selected clusters are selected for survey. Extra-large holdings will be in the sample by probability 1. That is, all clusters of extra-large holdings and all extra-large holdings from these clusters fall into sample. Primary sampling unit in the rest of the strata is the cluster. The same number of holdings will be interviewed in all the selected clusters of a stratum. Specifically, in small holding strata, 12 holdings will be interviewed in each selected cluster. This number is 8 for medium-sized strata and 4 for large strata. In each stratum the number of clusters that have to be selected is calculated by dividing the number of holdings to be selected in the stratum by the number of holdings to be interviewed in each cluster of the stratum. In each stratum selection of clusters is done by the PPS method (Probability Proportionally to Size). -The selection of holdings in each selected cluster is made using a random systematic sample.
• Rotational design: Survey has a panel design. Holdings, which will get into the sample, will stay there for three years. After this, they will be substituted by holdings from the same stratum. The database lists 943 extra-large holdings. All of them will constantly participate in the survey. Their rotation group number will be "0". Of the remaining holdings each of them will belong to one of the three rotation groups. Holdings selected from the same cluster will fall in the same rotation group. Each rotation group will have more or less the same number of holdings. Each rotation group represents an independent random sample. When holdings change by rotation , holding from the sample will be substituted by the new one from the same cluster. If the cluster does not have enough holdings to make the full rotation, then the cluster is deemed exhausted and is substituted by a randomly selected cluster from the same stratum. Newly introduced holdings will belong to the same rotation group which its predecessor belonged to

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

Computer Assisted Personal Interview [capi]

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

Detailed information on structure, and sections of questionnaires used in the survey of agricultural holdings available in following link: https://www.geostat.ge/en/modules/categories/686/agriculture-holdings-surveys.

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

After the field work, cleaning and harmonization of all inquiries are established at the Geostat head office - logical and arithmetical inconsistencies, as well as non-typical and suspicious data are detected, checked and corrected. Verification of the data is performed by contacting the respondents by phone. If verification with respondent is impossible, different imputation methods are used. Finally, indicators are calculated using weighted data. The obtained results are compared with corresponding results of the previous periods. In case of significant differences, the possible causes are identified and analyzed.

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

In the PME survey 237 holdings were not responded to due to refusing to be interviewed or would not be found during the fieldwork despite its existence. It is about 4.0% of the total Sampled holdings 5,880 holdings involved in the sample.

摘要
---------------------------
生产方法与环境模块调查的样本设计基于当前农业经营户调查的样本，因此首先给出当前调查的设计。农业经营户调查以及生产方法与环境模块调查的主要目的是产生符合农业部门的官方指标。该调查允许编制有关作物和畜牧业统计，包括年度和永久作物信息、播种面积、年度作物平均产量、耕作实践及其与自然环境的关系、作物和畜牧业生产方法、信息服务的获取和使用、基础设施和社区资源等。统计表可通过以下链接获取：https://www.geostat.ge/en/modules/categories/196/agriculture。

生产方法与环境模块是农业经营户调查的一部分。一轮主要调查（基准年）包括五次调查：基准年的起始访谈在当年的1月至2月期间使用起始问卷进行。在此访谈中，识别样本经营户并记录该经营户截至1月1日的情况。I、II和III季度访谈在基准年相应季度的下一个月初通过季度问卷进行。基于这些调查，收集了相应季度的农业活动信息。最终访谈在基准年下一年的1月通过最终问卷进行。在此访谈中，收集了基准年第四季度的农业活动信息以及该经营户在整个基准年（从上一年1月1日至12月31日）的农业活动汇总信息。在所有五次访谈中，对相同的农业经营户（约12000个）进行访谈，这些经营户是通过从约642000个在格鲁吉亚运营的农业经营户中选出的两阶段分层集群随机抽样程序选出的。在第一阶段，选择集群（定居点）。在第二阶段，在选定的集群内选择经营户。

调查完全覆盖了格鲁吉亚的领土，不包括阿布哈兹自治共和国和茨欣瓦利地区的被占领土。每年根据轮换设计（3年一换）选择新的样本。具体而言，每年约4000个样本经营户中的12000个将被新的经营户所取代。样本经营户参与调查3年。大型农业经营户每年进行抽样，实现全覆盖。调查的统计单位是农业经营户（家庭经营户和农业企业）——它被定义为在单一管理下的农业生产经济单位，包括所有饲养的牲畜和全部用于农业生产的土地，不考虑所有权、法律形式或规模。在所有者（在家庭的情况下——家庭成员，在农业企业的情况下——董事或授权人）的监督下进行农业活动，所有者负责做出决策并承担与农业活动相关的所有经济风险和费用。超过270名访谈员参与调查工作。在家庭经营户中，使用计算机辅助个人访谈方法（CAPI）进行数据收集。在农业企业中，由企业的授权人员（受访者）自行填写电子（在线）问卷（CAWI）。访谈员和收集数据的初步控制由协调员执行。他们的工作区域涵盖几个市政区。协调员的功能还包括就调查相关的方法论和技术问题为农业企业提供咨询。生产方法与环境模块的实地工作于2022年5月5日至5月20日进行。200名现场工作人员参与了调查，其中22名为现场监督员。总共选定了5,880个农业经营户进行PME调查。这些是持续参与调查的超大型农场以及自2019年以来一直参与调查的第三轮农场。目前，有943个超大型农场和3,899个第三轮农场参与调查。因此，我们有过去三年总共4,842个农场数据。其余的经营户将从第一轮集群中选出，这些集群已经进行了两年的访谈。具体而言，在每个层中使用简单随机抽样大约选择每个工作集群的30%。这将给我们大约1,038个农场。总共将选择约5,880个农场。

地理覆盖范围
---------------------------
整个国家（格鲁吉亚），不包括被占领土（阿布哈兹和茨欣瓦利地区）

分析单位
---------------------------
农业经营户——在单一管理下的农业生产经济单位，包括所有饲养的牲畜和全部用于农业生产的土地，不考虑所有权、法律形式或规模，其中在所有者的监督下进行农业活动，所有者负责做出决策并承担与农业活动相关的所有经济风险和费用。

总体
---------------------------
调查抽样框架包括约642000个在国家运营的农业经营户（家庭和农业企业）。2014年农业普查是抽样框架的主要来源。抽样框架根据农业经营户调查、商业登记簿和不同的行政来源的结果进行永久性更新。

数据类型
---------------------------
样本调查数据 [ssd]

抽样程序
---------------------------
生产方法与环境模块调查的样本设计基于当前农业经营户调查的样本设计，因此首先给出当前调查的设计。
• 自2016年以来抽样框架的主要来源——2014年农业普查；
• 抽样框架包含642000个经营户——样本量12000（1.9%）；
• 样本设计：两阶段分层集群随机抽样；——第一阶段——选择集群（定居点）；——第二阶段——在选定的集群内选择经营户；
• 每年根据轮换设计选择新的样本；——每年选出的1/3的经营户（4000）在前一年被取代（样本经营户参与调查3年）；
• 极大型农业经营户每年进行抽样，实现全覆盖；
• 更新抽样框架的附加来源：样本农业经营户调查、统计商业登记簿、MEPA中存在的行政数据（大型农业经营户）；主要指标的主观抽样误差不超过国家水平的5%和地区水平的10%。

生产方法与环境模块调查的样本设计：
• 样本设计：使用两阶段集群抽样进行调查。样本在每个层中分别形成。首先，在每个层中选定集群，然后从选定的集群中选择经营户进行调查。极大型经营户将以1的概率进入样本。即，所有极大型经营户的集群以及这些集群中的所有极大型经营户都将进入样本。除其他层以外的其他层的主要抽样单位是集群。每个选定的层中的集群数量将与该层中每个集群中要访谈的经营户数量相同。具体而言，在小型经营户层中，每个选定的集群将访谈12个经营户。这个数字在中等规模层中为8，在大规模层中为4。在每个层中，要选定的集群数量是通过将层中要选定的经营户数量除以层中每个集群中要访谈的经营户数量来计算的。在每个层中，通过PPS方法（按规模成比例的概率抽样）进行集群的选择。——在每个选定的集群中，通过随机系统抽样选择经营户。
• 轮换设计：调查具有面板设计。将进入样本的经营户将保持三年。在此之后，它们将被来自同一层的经营户所取代。数据库列出了943个极大型经营户。它们都将持续参与调查。它们的轮换组号将为“0”。其余的经营户中的每一个都将属于三个轮换组中的一个。来自同一集群的经营户将属于同一轮换组。每个轮换组将有或多或少相同数量的经营户。每个轮换组代表一个独立的随机样本。当经营户通过轮换而变化时，样本中的经营户将被来自同一集群的新经营户所取代。如果一个集群没有足够的经营户来完成完整的轮换，那么该集群被认为已经耗尽，并由来自同一层的随机选定的集群所取代。新引入的经营户将属于其前任所属的同一轮换组。

数据收集方式
---------------------------
计算机辅助个人访谈 [capi]

研究工具
---------------------------
有关在农业经营户调查中使用的问卷的结构和部分的详细信息可在以下链接中找到：https://www.geostat.ge/en/modules/categories/686/agriculture-holdings-surveys。

数据清理操作
---------------------------
在实地工作之后，Geostat总部建立了所有调查的清理和协调——检测、检查和纠正逻辑和算术不一致性、非典型和可疑数据。通过电话联系受访者进行数据验证。如果无法与受访者进行验证，则使用不同的插补方法。最后，使用加权数据计算指标。获得的结果与前期相应的结果进行比较。在出现重大差异的情况下，确定可能的原因并进行分析。

响应率
---------------------------
在PME调查中，由于拒绝接受访谈或尽管存在但无法在实地工作中找到，有237个经营户未响应。这大约占总样本量5,880个样本经营户的4.0%。