Quarterly Labour Force Survey 2012 - St. Lucia

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

The Labour Force Survey aims to collect information on the supply side of the labour market. It provides information on the extent of available and unused labour time and on relationships between employment and income. Thus, the data collected can be used for:

Macro-economic monitoring:- from an economic point of view, a main objective of collecting data on the economically active population is to provide basic information on the size and structure of a country's workforce. The unemployment rate in particular is widely used as an overall indicator of the current performance of a country's economy.

Human resources development: The economy is changing all the time. In order to meet the needs of the changing economy, people need to be trained. These areas of training must therefore be identified.

Employment policies: For an economy to work at its maximum potential, all persons wanting to have work should have jobs. Some persons may wish to have full-time jobs, and can only find part-time work. We need to know what proportion of the labour force these people represent in order to assess the social effects of government employment policies.

Income Support and social programmes: For the majority of people, employment income is their main means of support. People need not only jobs, but more importantly, productive jobs in order to receive reasonable incomes. We need to know what levels of income are being earned by different groups of persons.

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

National Coverage

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

- Households;
- Individuals.

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

The survey covered all de jure non-institutional household members (usual residents), it focuses on the employment, unemployment and current activity or inactivity status of all persons aged 15 years and over resident in the household.

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

Sample survey data [ssd]

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

Every quarter (three months) approximately 1,000 households are interviewed, there is a one third overlap between the households interviewed between each round of the survey.

The Multi-Stage sampling procedure developed for the St. Lucia MS (Master Sample) Frame is used for the execution of the labour force survey:

The two stage process of sample selection in the ST. LUCIA MS entails the selection of the PSUs within the districts. This is followed by the systematic selection of the cluster of households or USU (Ultimate Sampling Units) within the selected PSUs. The two stages in the design is elaborated as follows:

a. In the first stage, a sampling frame is constructed consisting of all of the enumeration districts from the census of 2001. The size of each enumeration district is measured in units of clusters of households. In the case of the ST. LUCIA MS, approximately seven or eight households were allocated per cluster. The clusters which are allocated to the EDs all have an equal probability of selection within the specified geographic domain in which they are allocated. In addition, the number of clusters allocated to an ED is a measure of the size of the ED. Clusters, therefore ensure the selection of EDs or Primary Sampling Units with probability proportional to the size of the ED. The ST. LUCIA MS frame consists of nine sub-samples / replicates, with each replicate selected with a probability of (1 / (16 * 9)) or 1 / 144.

b. In the second stage a non-compact cluster of households is selected within the selected PSU using systematic random sampling. There are three elements to the selection of this non-compact cluster. Firstly, there is the sample interval, which is a measure of the size of the ED in terms of the total number of households it contains. The larger the ED or PSU the larger will be the sample interval assigned and consequently the larger will be the number of clusters assigned to the ED. This approach ensures that the total number of households selected in any selected ED is approximately the same. In the case of the "Castries" in the ST. LUCIA MS frame the approximate number is five (5). Secondly, the random start is determined by use of a random number generator. With a Microsoft EXCEL spreadsheet the formulae takes the following form, =ROUND(RAND()*E1,0)+1, where E1 is the cell containing the sample interval (or total number of clusters assigned) RAND() is the function which generates the random number. The round() function is used to round the result to the nearest whole number. The third element of choosing the non compact cluster is a combination of the above. A random number (r) is choosen between 1 and the sample interval value, I, inclusive, then to this number is added the sample interval for the full list of households within the primary sample unit. Thus, the list of selected households would be r, r + I, r + 2I, r + 3I, r + 4I,……, r + (n - 1)I, where n is the cluster size assigned to the district, in the case of Castries n is five.

A. Size of the Sample

As has been explained before the decision to use a sampling fraction of 1 : 16 and to assign nine replicates to each District (the geographic domain) was based on the need to take advantage of the small size of the countries covered by this MECOVI project. This was done by increasing the "spread" of the sample across EDs and as a result improving the precision of the estimates which can be obtained from it. In addition, attention was paid to ensuring that were the CSO of ST. LUCIA to consider developing further its Integrated Household Survey Programme, the ground work would have been laid through this Master Sample Frame design for periodic, ad hoc or continuous sample surveys. The achievement of this objective has already been demonstrated through the use of this Sample Frame in the conduct of St. Lucia's continuous Labour Force Survey.

Therefore for any one sub-sample given that there are nine, the sampling fraction is 1 / 16 by 1 / 9 or 1 / 144. If a periodic, ad hoc or quarterly survey included the use of three replicates then the sampling fraction for these three replicates would be 3 / 144 or 1 /16 by 3 / 9. In both cases the resultant sampling fraction is the product of the sampling probability for the Master Sampling frame and the probability of selection of a specific number of replicates.

B. Master Sample Domains of Study and Stratification

1. Domains of Study:

The Master Sample frame was subdivided into eleven areas for the purpose of the provision of estimates from samples selected from this frame. The following list of the ten domains or sub-populations is based on the Districts which formed the basis for the collection of information on the population in the 2001 Census.

The total number of PSUs in the ST. LUCIA MS is 401, a breakdown of the number of PSUs by District is shown in the table above. The average size of the PSUs was 118 approximately with a standard deviation of approximately 47. This configuration does not in the near term present a major problem for sample implementation, since the EDs/PSUs size does not exceed 100 by too great an extent, in addition, while consideration must be given to splitting EDs which have grown in size to over 200, there are not as exist in the case of St. Vincent and the Grenadines a significant number of excessively large EDs. Continuous maintenance of this situation is required and can be done by splitting all EDs over 200 in size into smaller ones of approximate size 100. The main objective of controlling the size of the PSUs, is to reduce variability and thereby improve the precision of estimates from the sample. The more equal the sizes of the PSUs the more likely the variance of characteristics between PSUs will be minimized and inversely the precision of the samples derived from the estimates from the Master Sample Frame increased.

2. Stratification

As shown in the table above each of the domains of study was stratified according to specific criteria. In the more urban domains the criteria used was the percentage of Managers, professional, sub-professionals in the population. The PSUs or EDs were therefore arranged in descending order of the proportion of this group in the population of the ED. In the rural domains the PSUs were arranged in descending order of the proportion of agriculture workers in the population of the ED. In the case of Canaries and Anse-la-Raye, the sizes of the populations in these domains mandated a joining of the two to allow for the creation of a large enough domain for reporting purposes.

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

Face-to-face [f2f]

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

The questionnaire is administered to all members of the household. Questions 1 through 6 are to be completed for all members of the household, these questions cover age, sex, relation to head of household, country of birth etc. All subsequent questions refer to persons 15 year of age and older. The questionnaire is divided into five parts:

PART 1:For all members of the household (regardless of age) - Demographic and emigration questions
PART 2: To be completed for persons 15 years and older - Education, Training, activities during the reference week or month, working at a job, on vacation, methods of seeking work, availability for employment
PART 3: For persons employed during the reference week - Number of actual hours of work, number of usual hours of work, seeking additional work, status in employment, industry and occupation of employment
PART 3A: For persons holding more than one job during the reference week - Number of actual hours of work, number of usual hours of work, seeking additional work, status in employment, industry and occupation of employment
PART 4: For persons unemployed during the reference week - Lenght of unemployment, means of support, occupation in last job
PART 5: For persons not in the labour force during the reference week, first seekers, means of support, intention to join the labour force in the future.

The questionnaire used to conduct the labour force survey is provided as an external resource.

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

Data editing takes place at a number of stages throughout the processing, including:
a) Office editing and coding when the questionnaires are delivered to the Statistics Office by the enumerators
b) During data verification of the scanned questionnaire
c) Structure checking and consistency editing and completeness is performed after data capture to MS SQL Server
d) Secondary editing is completed in CSPRO where a software program for this purpose have been written
e) Final production of SPSS data files after all editing is completed in CSPRO.

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

On average the response rate on the Saint Lucia Labour force survey is over 85%.

Sampling error estimates
---------------------------

Estimates from a sample survey are affected by two types of errors: 1) non-sampling errors and 2) sampling errors. Non-sampling errors are the results of mistakes made in the implementation of data collection and data processing. Numerous efforts are made during implementation of the quarterly labour force survey to minimize this type of error, however, non-sampling errors are impossible to avoid and difficult to evaluate statistically.

If the sample of respondents had been a simple random sample, it would have been possible to use straightforward formulae for calculating sampling errors. However, the quarterly labour force sample survey is the result of a multi-stage stratified design, and consequently needs to use more complex formulae. The SPSS complex samples module is being used to calculate sampling errors for the quarterly labour force survey for St Lucia. This module uses the Taylor linearization method of variance estimation for survey estimates that are means or proportions. This method is documented in the SPSS file CSDescriptives.pdf found under the Help, Algorithms options in SPSS.

Sampling errors have been calculated for a select set of statistics (all of which are proportions due to the limitations of the Taylor linearization method) for the national sample. For each statistic for which sample errors are computed, the estimate, its standard error, the coefficient of variation (or relative error -- the ratio between the standard error and the estimate), the design effect, and the square root design effect (DEFT -- the ratio between the standard error using the given sample design and the standard error that would result if a simple random sample had been used), as well as the 95 percent confidence intervals (+/-2 standard errors).

Details of the sampling errors are presented in the sampling errors table presented in the external resources.

摘要
---------------------------
劳动力市场调查旨在收集劳动力市场的供给方信息。它提供了关于可用和未使用的劳动时间范围以及就业与收入之间关系的信息。因此，收集到的数据可用于以下方面：

宏观经济监测：从经济角度来看，收集经济活跃人口数据的根本目标是为一个国家的劳动力规模和结构提供基本信息。特别是，失业率被广泛用作衡量一个国家当前经济表现的整体指标。

人力资源发展：经济时刻在变化。为了满足不断变化的经济需求，人们需要接受培训。因此，必须确定这些培训领域。

就业政策：为了使经济达到其最大潜力，所有希望就业的人士都应找到工作。有些人可能希望全职工作，但只能找到兼职工作。我们需要知道这些人在劳动力中所占的比例，以便评估政府就业政策的社会影响。

收入支持和社会福利计划：对于大多数人来说，就业收入是他们主要的收入来源。人们不仅需要工作，更需要能够获得合理收入的生产性工作。我们需要了解不同群体人士所赚取的收入水平。

地理覆盖范围
---------------------------
全国覆盖

分析单位
---------------------------
- 家庭；
- 个人。

总体
---------------------------
调查涵盖了所有法定非机构家庭成员（常住居民），重点关注所有15岁及以上居民的家庭中的就业、失业和当前活动或不活动状态。

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

抽样程序
---------------------------
每季度（三个月）大约调查1000户家庭，每次调查之间的家庭存在三分之一的重叠。

用于圣卢西亚MS（主样本）框架执行的劳动力市场调查采用了以下多阶段抽样程序：

在圣卢西亚MS中，样本选择的两阶段过程包括在地区内选择PSU（初级抽样单位）。这随后是由选定的PSU内的家庭集群或USU（最终抽样单位）的系统性选择。设计中的两个阶段如下所述：

a. 在第一阶段，构建了一个包含2001年人口普查中所有人口普查区的抽样框架。每个人口普查区的规模以家庭集群为单位进行衡量。在圣卢西亚MS的情况下，每个集群分配了大约七到八个家庭。分配给ED的所有集群在分配的指定地理区域内都有相同的选择概率。此外，分配给ED的集群数量是ED规模的一个衡量标准。因此，集群确保了与ED或初级抽样单位的规模成比例的选择概率。圣卢西亚MS框架由九个子样本/重复样本组成，每个重复样本的选择概率为（1/（16*9））或1/144。

b. 在第二阶段，使用系统性随机抽样在选定的PSU内选择非紧凑的家庭集群。选择这个非紧凑集群有三个要素。首先，有一个样本间隔，这是衡量ED规模的一个指标，即它包含的总家庭数量。ED或PSU越大，分配的样本间隔就越大，因此分配给ED的集群数量也越大。这种方法确保了在任何选定的ED中选定的家庭总数大致相同。在圣卢西亚MS框架的“卡斯特里”的情况下，大约是五（5）。其次，随机起始点是通过使用随机数发生器确定的。使用Microsoft EXCEL电子表格时，公式如下，=ROUND(RAND()*E1,0)+1，其中E1是包含样本间隔（或分配的集群总数）的单元格 RAND()是生成随机数的函数。round()函数用于将结果四舍五入到最接近的整数。选择非紧凑集群的第三个要素是上述组合。选择一个介于1和样本间隔值I之间的随机数（r），然后将其与初级抽样单位内所有家庭的完整列表中的样本间隔相加。因此，选定的家庭列表将是r，r + I，r + 2I，r + 3I，r + 4I，……，r + (n - 1)I，其中n是分配给地区的集群大小，在卡斯特里的情况下，n是五。

A. 样本量的大小

如前所述，使用1:16的抽样比例，并为每个地区（地理区域）分配九个重复样本的决定是基于利用本MECOVI项目涵盖的国家的规模较小的需要。这是通过增加样本在ED上的“分布”来实现的，从而提高了从样本中获得的估计值的精度。此外，还注意确保如果圣卢西亚的CSO（国家统计局）考虑进一步发展其综合家庭调查计划，则通过本主样本框架设计将为定期、临时或连续样本调查奠定基础。这一目标已经通过使用本样本框架进行圣卢西亚连续劳动力市场调查的执行得到了证明。

因此，对于任何给定的子样本，考虑到有九个，抽样比例是1/16乘以1/9或1/144。如果定期、临时或季度调查包括使用三个重复样本，那么这三个重复样本的抽样比例将是3/144或1/16乘以3/9。在这两种情况下，结果抽样比例是主抽样框架抽样概率与选择特定数量重复样本概率的乘积。

B. 研究领域和分层

1. 研究领域：

主样本框架被划分为十一个区域，以从该框架中选择的样本提供估计值。以下基于2001年人口普查形成信息收集基础的十个领域或子群体列表如下。

圣卢西亚MS中的PSU总数为401，按地区划分的PSU数量在上述表中显示。PSU的平均大小约为118，标准差约为47。这种配置在短期内不会对样本实施造成重大问题，因为ED/PSU的规模并不超过100太多。此外，虽然必须考虑到将增长到超过200的ED分割成更小的单位，但与圣文森特和格林纳丁斯的情况不同，没有存在大量过大ED的情况。需要持续维护这种情况，并且可以通过将所有超过200大小的ED分割成大小约为100的小单位来实现。控制PSU大小的主要目标是减少变异性，从而提高样本估计值的精度。PSU的规模越均匀，PSU之间特征差异的方差就越有可能最小化，反之，从主样本框架估计值中得出的样本的精度就越高。

2. 分层：

如上表所示，每个研究领域都根据特定标准进行分层。在更城市化的领域，使用的标准是人口中经理、专业人员和半专业人员所占的百分比。因此，PSU或ED按照在ED人口中这一群体所占比例的降序排列。在乡村领域，PSU按照在ED人口中农业工人所占比例的降序排列。在卡尼和安斯-拉雷的情况下，这些领域中的人口规模要求将这两个领域合并，以便为报告目的创建足够大的领域。

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

研究工具
---------------------------
问卷由家庭的所有成员填写。第1至6题应由家庭的所有成员完成，这些问题涵盖年龄、性别、与户主的关系、出生国等。所有后续问题均针对15岁及以上的人。问卷分为五个部分：

第一部分：针对所有家庭成员（无论年龄）- 人口和移民问题

第二部分：针对15岁及以上的人 - 教育、培训、参考周或月期间的活动、工作、休假、寻找工作的方式、就业可用性

第三部分：针对参考周内就业的人员 - 实际工作时间数、通常工作时间数、寻找额外工作、就业状态、行业和就业职业

第三部分A：针对参考周内持有多个工作的人员 - 实际工作时间数、通常工作时间数、寻找额外工作、就业状态、行业和就业职业

第四部分：针对参考周内失业的人员 - 失业长度、支持手段、最后工作的职业

第五部分：针对参考周内不在劳动力中的人员，首次求职者、支持手段、未来加入劳动力的意愿。

用于进行劳动力市场调查的问卷作为外部资源提供。

数据清理操作
---------------------------
数据编辑在数据处理的多个阶段进行，包括：

a) 当调查员将问卷递送到统计局时进行办公室编辑和编码

b) 在扫描问卷的数据验证期间

c) 在数据捕获到MS SQL Server后进行结构检查和一致性编辑，并执行完整性检查

d) 在CSPRO中完成二级编辑，其中编写了为此目的的软件程序

e) 在CSPRO中所有编辑完成后，生成SPSS数据文件。

回应率
---------------------------
圣卢西亚劳动力市场调查的平均回应率超过85%。

抽样误差估计
---------------------------
样本调查的估计值受两种类型的误差影响：1) 非抽样误差和2) 抽样误差。非抽样误差是数据收集和数据处理实施中错误的结果。在实施季度劳动力市场调查的过程中，付出了大量努力来最大限度地减少此类错误，然而，非抽样误差是不可避免的，并且难以进行统计评估。

如果受访者样本是一个简单随机样本，那么就可以使用简单的公式来计算抽样误差。然而，季度劳动力市场样本调查是多层次分层设计的产物，因此需要使用更复杂的公式。正在使用SPSS复杂样本模块来计算圣卢西亚季度劳动力市场调查的抽样误差。此模块使用泰勒线性化方法进行方差估计，用于调查估计值是平均值或比例。这种方法在SPSS文件CSDescriptives.pdf中有记录，该文件位于SPSS“帮助”选项下的“算法”选项下。

为选定的一组统计数据（所有这些都是由于泰勒线性化方法的限制而成为比例）计算了抽样误差。对于每个计算抽样误差的统计数据，估计值、其标准误差、变异系数（或相对误差--标准误差与估计值的比率）、设计效应以及平方根设计效应（DEFT--使用给定样本设计计算的标准误差与如果使用简单随机样本将产生标准误差之间的比率），以及95%置信区间（±2标准误差）。

抽样误差的详细信息在抽样误差表中提供，该表作为外部资源。