Syntax to support users
Details are provided below of a range of syntax files to support users in their analysis of WERS data.
1. Full WERS Research Team syntax
Much of the Stata syntax created by the WERS Research Team in their primary analysis of the 2011 survey has been made available to those engaged in secondary analysis of the data.
This syntax will be useful to anyone wishing to extend the primary analysis in some way, as it enables users to precisely replicate the derived variables created by the WERS Research Team. This helps users to avoid uncertainty about how the Research Team has used the data, and to avoid unecessary duplication of effort.
The syntax is in Stata format (only) and has been made available in two parts:
1. A suite of syntax files which compile data from the 2004 and 2011 surveys and construct a wide range of derived variables - thus creating the core Stata dataset used by the 2011 WERS Research Team in all of their primary analysis.
2. A suite of syntax files which use the dataset cited above to generate all of the estimates reported in the 2011 WERS First Findings booklet.
Both sets of syntax files are provided as part of the WERS data deposit that can be accessed via the UK Data Service. The syntax files were added in the Fifth Edition of the deposited data. The syntax files come with documentation which explains their structure and content.
2. Syntax to create a workplace (MQ) data file for 2004 and 2011
* Includes syntax needed to create a 2004-2011 MQ panel dataset *
To support users wishing to conduct longitudinal analysis of the workplace data (only) in WERS 2004 and 2011, we have compiled a short syntax file which combines MQ data from the two surveys into a single datafile. This allows for easy comparison of the incidence of workplace practices across the two years without the need to compile the full WERS Research Team dataset (described above).
Panel workplaces (those interviewed in both 2004 and 2011) are necessarily included in this dataset. The syntax thus provides a means of compiling an MQ panel dataset for 2004-2011. An indicator variable (npanel==1) allows users to extract this sub-sample from the full dataset.
The syntax file can be downloaded from the UK Data Service catalogue page for WERS 2011 (scroll to the bottom of the page for the section headed 'Syntax/code').
3. Syntax to create weights for multi-level modelling
The linked SEQ-MQ data from WERS 2011 provides opportunities for multi-level analysis to explore the importance of employee and workplace-level predictors. However, multi-level regression estimators (such as Stata's -mixed- command) require the user to specific separate weights for each level of the sample design. The WERS 2011 SEQ dataset contains a single weight (seqwtnrc_apr13) which takes account of sample selection at both the workplace level (Level 2) and employee level (Level 1). The syntax below decomposes this combined SEQ weight variable (seqwtnrc_apr13) into two components:
- a Level 1 weight which takes account of the employee's probability of selection, conditional on their workplace having participated in the SEQ
- a Level 2 weight which takes account of the workplace's probability of participating in the SEQ.
The syntax file can be downloaded from the UK Data Service catalogue page for WERS 2011 (scroll to the bottom of the page for the section headed 'Syntax/code').
We have also written a background note to describe the derivation process and associated caveats.
Download the background note (PDF file)
4. Syntax to code occupational groups in the MQ and SEQ
Users commonly ask how to identify:
- the largest occupational group within the workplace (a frequent reference point throughout the Management Questionnaire)
- the occupational group of employee respondents to the SEQ.
Identifying the largest occupational group in the workplace:
The largest occupational group in the workplace can be identified using the variable 'nhiocc', which is provided on the MQ dataset deposited with the UK Data Service. The variable is coded to the Major Group level of the SOC(2010) classification. The accompanying variable 'nhiocc1' stores the number of employees within the workplace who sit within this SOC(2010) Major Group. These data are collected in the Employee Profile Questionnaire.
The variable 'nsoc2010' contains the detailed occupation that is collected at question ZSOCDESC in the Management Questionnaire. These data are coded to the Unit Group level of SOC(2010). The first digit of 'nsoc2010' does not always match 'nhiocc', despite efforts to resolve inconsistencies with respondents during and after fieldwork.
Identifying the occupational group of employee respondents to the SEQ:
The occupation of each employee respondent to the SEQ is held in the variable 'xsoc2010'. The data are coded to the Unit Group level of SOC(2010). We have written a small syntax file which generates a SOC(2010) Major Group (one-digit) code from this four-digit classification.
The syntax file can be downloaded from the UK Data Service catalogue page for WERS 2011 (scroll to the bottom of the page for the section headed 'Syntax/code').
The 2011 SEQ dataset also also contains 'xsoc2000' which is the employee's occupation coded to the SOC(2000) classification. This variable has been included in the 2011 dataset in order to promote consistency with the WERS 2004 SEQ dataset, which uses SOC(2000) codes.
5. Syntax to create a harmonised WERS/REPONSE dataset
In a recent Leverhulme-funded project (RPG2013-399), some members of the 2011 WERS Research Team colloborated with researchers involved in the French equivalent survey (the 2011 REPONSE) to create a single, harmonised dataset containing data from the Management and Employee Questionnaires of the 2004 and 2011 WERS and REPONSE surveys.
The Stata syntax files to derive this harmonised WERS/REPONSE dataset can be downloaded from the project website hosted by NIESR (scroll to the bottom to the section headed 'Documentation to permit further comparative analysis').
The syntax files are accompanied by translated questionnaires and documentation summarising the methodologies used in the surveys.
