Georgia Tobacco Survey 2017-2019 - Georgia

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

The Georgia Tobacco Survey (GTS) is a panel survery that collects data from smokers (and former smokers in follow-up waves) on tobacco usage in the Eastern European country of Georgia. The survey covers five regions within Georgia (Tbilisi, Kutaisi, Zugdidi, Akhaltsikhe and Gori) and consists of three waves: GTS 2017 (wave 1), GTS 2018 (wave 2) and GTS 2019 (wave3). Wave 1 was conducted in November 2017 as a response to the passing of a new tobacco contral law in 2017. Wave 2 was conducted in December 2018 after the tobacco control law was enacted in May 2018. Wave 3 was conducted in May 2019 following an increase in cigarette excise tax in January 2019. Tobacco taxes are critical in reducing tobacco consumption and thereby improving public health. However, the tobacco industry claims that a tax increase will increase the size of the illicit tobacco market. Therefore, the GTS can be used to assess the extent of the illicit tobacco market in the country.

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

The survey has national coverage, although it is worth noting that data was collected in 5 of Georgia's 12 regions (Tbilisi, Kutaisi, Zugdidi, Akhaltsikhe and Gori). These five regions represented 41.2% of the total Georgian population and were chosen based on geographical diversity and/or likely illicit trade hot spots given their proximity to borders.

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

Households and individuals

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

The Georgia Tobacco Survey covers all adults in households (aged 18 and older).

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

Sample survey data

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

The sampling began with urban and rural stratification, followed by the selection of primary sampling units (census units in urban strata and villages in rural strata) proportional to population size. Secondary sampling units (households) were then selected using the 'random step method'. If no smokers resided in the selected household, a short general survey was administered to the first adult person answering the door. If the household had a smoker, all smokers were listed and one was randomly selected for the interview. If the selected respondent refused or was not home during any of the three interview attempts, another smoking resident was selected. If no smoking residents were available for interview, the household was recorded as 'non-responding' and the interviewer walked past five households on average (the number varied slightly for urban and rural strata) and selected a replacement household.

Surveyors completed 2997 interviews in wave 1. Of those, 1765 people were successfully re-interviewed in wave 2. These were supplemented with new households (also selected using the 'random step method') which brought the total people interviewed in wave 2 to 3040. In wave 3, there were 1578 people interviewed in both waves 1 and 2 (our panel), 240 people interviewed in wave 1 but not in wave 2, 1036 people only interviewed in wave 2, and 335 people added in wave 3 for the first time. As such, there were 3189 people interviewed in wave 3. For quality control, an independent controller crosschecked a randomly selected 10% of completed interviews in each wave and no major errors were identified.

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

Face-to-face [f2f]

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

The surveys conducted in all three waves were similar in format, with a few extra questions added in wave 3 to determine any behavioural changes resulting from the new law and tax increase.

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

The data received by DataFirst had been cleaned, merged, and processed for analysis before arriving at DataFirst. Further to this, significant data cleaning was undertken by DataFirst. DataFirst cleaned unreadable characters, empty variables and redundant variables out of the data. The data was also anonymised. The syntax files for all of this cleaning is available with the data. Additionally, significant cleaning of the geographic variables was necessary (see data quality notes).

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

The total attrition in our panel was 52.6% with attrition varying significantly by region. The highest rate was in Tbilisi (62.7%) followed by Kutaisi (52.2%), Akhaltsikhe (47.4%), Gori (31.9%) and Zugdidi (27.4%). Within each region, there was no significant difference in the likelihood of attrition by gender, employment, the probability of showing a pack or the probability of owning an illicit pack. Given the attrition patterns, the analysis was done by region wherever possible.

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

There are a few data quality issues at present:

Time: In the current version 1, there are no usable time variables from any of the waves. There are time variables in wave 1 but it is not clear the units these are measured in - they do not correspond to the usual epoch values one would expect.

Geographic: The locational information was not asked consistently in the questionnaires and the resulting data required some fixing. In urban settings, respondents tended to provide region, city, and city district, whereas in rural areas they provided region, district and village. Furhter, for anonymization village had to be redacted. As such, in urban settings there are 3 geographic levels in the cleaned data but only 2 for rural. Each unique case was also checked on the map and inconsistencies were attempted to be fixed. In some cases inconsistent were unfixable and were then left unchaged.

Weight: At present it is not clear what population totals the weights are calibrated to, or what the data represent when weighted.

摘要
---------------------------
乔治亚烟草调查（GTS）是一项面板调查，旨在收集来自乔治亚东部欧洲国家吸烟者（以及在后续波次中的前吸烟者）的烟草使用数据。该调查覆盖了乔治亚境内的五个地区（第比利斯、库塔伊西、祖格迪迪、阿哈尔齐赫和戈里），并分为三个波次：GTS 2017（波次1）、GTS 2018（波次2）和GTS 2019（波次3）。波次1于2017年11月进行，作为对2017年通过的新的烟草控制法的回应。波次2于2018年5月烟草控制法实施后于2018年12月进行。波次3在2019年1月烟草消费税增加后于2019年5月进行。烟草税对于减少烟草消费并从而改善公共健康至关重要。然而，烟草行业声称，税收增加将扩大非法烟草市场的规模。因此，GTS可用于评估该国非法烟草市场的规模。
地理覆盖范围
---------------------------
该调查具有全国覆盖范围，尽管值得注意的是，数据是在乔治亚的12个地区中的5个地区（第比利斯、库塔伊西、祖格迪迪、阿哈尔齐赫和戈里）收集的。这五个地区代表了乔治亚总人口的41.2%，其选择基于地理多样性以及/或可能成为非法贸易热点地区的可能性，鉴于其靠近边境。
分析单位
---------------------------
家庭和个人
总体
---------------------------
乔治亚烟草调查涵盖了所有家庭中的成年人（18岁及以上）。
数据类型
---------------------------
样本调查数据
抽样程序
---------------------------
抽样始于城市和农村分层，然后根据人口规模选择一级抽样单位（城市层中的普查单元和农村层中的村庄）。然后使用“随机步进法”选择二级抽样单位（家庭）。如果所选家庭中没有居住的吸烟者，则对第一个开门的成年人进行简短的通用调查。如果家庭中有吸烟者，则列出所有吸烟者，并随机选择一人进行访谈。如果所选受访者拒绝或在任何三次访谈尝试期间不在家，则选择另一位吸烟居民。如果没有可供访谈的吸烟居民，则该家庭被记录为“未响应”，调查员平均走过五个家庭（城市和农村层略有差异）并选择一个替代家庭。
调查员在波次1中完成了2997次访谈。其中，1765人在波次2中成功接受了重新访谈。此外，还补充了新的家庭（也使用“随机步进法”选择），使波次2中接受访谈的总人数达到3040人。在波次3中，有1578人在波次1和波次2中接受过访谈（我们的面板），240人在波次1中接受过访谈但不在波次2中，1036人只在波次2中接受过访谈，以及335人在波次3中首次接受访谈。因此，波次3中有3189人接受访谈。为了质量控制，每个波次随机抽查了10%的完成访谈，未发现重大错误。
数据收集方式
---------------------------
面对面[f2f]
研究工具
---------------------------
在所有三个波次中进行的调查在格式上相似，但在波次3中增加了几个额外的问题，以确定新法律和税收增加导致的任何行为变化。
数据清洗操作
---------------------------
收到的数据已由DataFirst进行清洗、合并和预处理，以便进行分析。此外，DataFirst还进行了重大的数据清洗。DataFirst从数据中清除了不可读字符、空变量和冗余变量。数据还被匿名化。所有这些清洗的语法文件都与数据一起提供。此外，对地理变量进行了重大的清洗（见数据质量说明）。
响应率
---------------------------
我们的面板中的总体流失率为52.6%，流失率在不同地区之间存在显著差异。最高的是第比利斯（62.7%），其次是库塔伊西（52.2%）、阿哈尔齐赫（47.4%）、戈里（31.9%）和祖格迪迪（27.4%）。在每个地区内，性别、就业、展示烟盒的概率或拥有非法烟盒的概率在流失的可能性上均无显著差异。鉴于流失模式，尽可能按地区进行数据分析。
数据评估
---------------------------
目前存在一些数据质量问题：
时间：在当前版本1中，没有任何一个波次有可用的时序变量。波次1中有时序变量，但它们测量的单位不清楚——它们不符合预期的通常纪元值。
地理：定位信息在问卷中未得到一致询问，导致的数据需要一些修正。在城市环境中，受访者倾向于提供地区、城市和城市区，而在农村地区，他们提供地区、区和村庄。此外，为了匿名化，村庄必须被删除。因此，在城市环境中，清洗后的数据中有3个地理层级，但在农村地区只有2个。每个唯一案例也都已在地图上进行了检查，并尝试修正不一致之处。在某些情况下，不一致之处无法修复，因此被保留原样。
权重：目前尚不清楚权重校准的总体人口总数是什么，或加权数据代表什么。